Bump version for 0.15.0

Updated README with info for Rails <7 usage

.github/workflows/ci.yml

@@ -1,5 +1,9 @@
 name: CI
-on: [push, pull_request]
+on: 
+  push:
+    branches:
+      - main
+  pull_request:
 jobs:
   tests:
     strategy:

CONTRIBUTING.md

@@ -27,7 +27,7 @@ Please keep the following guidelines in mind when opening a pull request:
 - Add tests for your changes, if possible.
 - Ensure that your changes don't break existing functionality.
 
-#### Commit message guidline
+#### Commit message guidelines
 
 A good commit message should describe what changed and why.
 

Dockerfile

@@ -14,14 +14,14 @@ COPY Gemfile Gemfile.lock mrsk.gemspec ./
 COPY lib/mrsk/version.rb /mrsk/lib/mrsk/version.rb
 
 # Install system dependencies
-RUN apk add --no-cache --update build-base git docker openrc \
+RUN apk add --no-cache --update build-base git docker openrc openssh-client-default \
     && rc-update add docker boot \
     && gem install bundler --version=2.4.3 \
     && bundle install
 
 # Copy the rest of our application code into the container.
 # We do this after bundle install, to avoid having to run bundle
-# everytime we do small fixes in the source code.
+# every time we do small fixes in the source code.
 COPY . .
 
 # Install the gem locally from the project folder
@@ -31,6 +31,10 @@ RUN gem build mrsk.gemspec && \
 # Set the working directory to /workdir
 WORKDIR /workdir
 
+# Tell git it's safe to access /workdir/.git even if
+# the directory is owned by a different user
+RUN git config --global --add safe.directory /workdir
+
 # Set the entrypoint to run the installed binary in /workdir
 # Example:  docker run -it -v "$PWD:/workdir" mrsk init
 ENTRYPOINT ["mrsk"]

Gemfile.lock

@@ -1,7 +1,7 @@
 PATH
   remote: .
   specs:
-    mrsk (0.11.0)
+    mrsk (0.15.0)
       activesupport (>= 7.0)
       bcrypt_pbkdf (~> 1.0)
       dotenv (~> 2.8)

README.md

@@ -44,30 +44,48 @@ Then edit your `.env` file to add your registry password as `MRSK_REGISTRY_PASSW
 Now you're ready to deploy to the servers:
 
 ```
-mrsk deploy
+mrsk setup
 ```
 
 This will:
 
 1. Connect to the servers over SSH (using root by default, authenticated by your ssh key)
-2. Install Docker on any server that might be missing it (using apt-get): root access is needed via ssh for this.
+2. Install Docker and curl on any server that might be missing it (using apt-get): root access is needed via ssh for this.
 3. Log into the registry both locally and remotely
 4. Build the image using the standard Dockerfile in the root of the application.
 5. Push the image to the registry.
 6. Pull the image from the registry onto the servers.
 7. Ensure Traefik is running and accepting traffic on port 80.
-8. Ensure your app responds with `200 OK` to `GET /up`.
+8. Ensure your app responds with `200 OK` to `GET /up` (you must have curl installed inside your app image!).
 9. Start a new container with the version of the app that matches the current git version hash.
 10. Stop the old container running the previous version of the app.
 11. Prune unused images and stopped containers to ensure servers don't fill up.
 
-Voila! All the servers are now serving the app on port 80. If you're just running a single server, you're ready to go. If you're running multiple servers, you need to put a load balancer in front of them.
+Voila! All the servers are now serving the app on port 80. If you're just running a single server, you're ready to go. If you're running multiple servers, you need to put a load balancer in front of them. For subsequent deploys, or if your servers already have Docker and curl installed, you can just run `mrsk deploy`.
+
+### Rails <7 usage
+
+MRSK is not needed to be in your application Gemfile to be used. However, if you want to guarantee specific MRSK version in your CI/CD workflows, you can create a separate Gemfile for MRSK, for example, `gemfile/mrsk.gemfile`:
+
+```ruby
+source 'https://rubygems.org'
+
+gem 'mrsk', '~> 0.14'
+```
+
+Bundle with `BUNDLE_GEMFILE=gemfiles/mrsk.gemfile bundle`.
+
+After this MRSK can be used for deployment:
+
+```sh
+BUNDLE_GEMFILE=gemfiles/mrsk.gemfile bundle exec mrsk deploy
+```
 
 ## Vision
 
 In the past decade+, there's been an explosion in commercial offerings that make deploying web apps easier. Heroku kicked it off with an incredible offering that stayed ahead of the competition seemingly forever. These days we have excellent alternatives like Fly.io and Render. And hosted Kubernetes is making things easier too on AWS, GCP, Digital Ocean, and elsewhere. But these are all offerings that have you renting computers in the cloud at a premium. If you want to run on your own hardware, or even just have a clear migration path to do so in the future, you need to carefully consider how locked in you get to these commercial platforms. Preferably before the bills swallow your business whole!
 
-MRSK seeks to bring the advance in ergonomics pioneered by these commercial offerings to deploying web apps anywhere. Whether that's low-cost cloud options without the managed-service markup from the likes of Digital Ocean, Hetzner, OVH, etc, or it's your own colocated bare metal. To MRSK, it's all the same. Feed the config file a list of IP addresses with vanilla Ubuntu servers that have seen no prep beyond an added SSH key, and you'll be running in literally minutes.
+MRSK seeks to bring the advance in ergonomics pioneered by these commercial offerings to deploying web apps anywhere. Whether that's low-cost cloud options without the managed-service markup from the likes of Digital Ocean, Hetzner, OVH, etc., or it's your own colocated bare metal. To MRSK, it's all the same. Feed the config file a list of IP addresses with vanilla Ubuntu servers that have seen no prep beyond an added SSH key, and you'll be running in literally minutes.
 
 This approach gives you enormous portability. You can have your web app deployed on several clouds at ease like this. Or you can buy the baseline with your own hardware, then deploy to a cloud before a big seasonal spike to get more capacity. When you're not locked into a single provider from a tooling perspective, there are a lot of compelling options available.
 
@@ -123,6 +141,8 @@ This template can safely be checked into git. Then everyone deploying the app ca
 
 If you need separate env variables for different destinations, you can set them with `.env.destination.erb` for the template, which will generate `.env.staging` when run with `mrsk envify -d staging`.
 
+Note: If you utilize biometrics with 1Password you can remove the `session_token` related parts in the example and just call `op read op://Vault/Docker Hub/password -n`.
+
 #### Bitwarden as a secret store
 
 If you are using open source secret store like bitwarden, you can create `.env.erb` as a template which looks up the secrets.
@@ -184,6 +204,19 @@ registry:
 
 A reference to secret `DOCKER_REGISTRY_TOKEN` will look for `ENV["DOCKER_REGISTRY_TOKEN"]` on the machine running MRSK.
 
+#### Using AWS ECR as the container registry
+
+AWS ECR's access token is only valid for 12hrs. In order to not have to manually regenerate the token every time, you can use ERB in the `deploy.yml` file to shell out to the `aws` cli command, and obtain the token:
+
+```yaml
+registry:
+  server: <your aws account id>.dkr.ecr.<your aws region id>.amazonaws.com
+  username: AWS
+  password: <%= %x(aws ecr get-login-password) %>
+```
+
+You will need to have the `aws` CLI installed locally for this to work.
+
 ### Using a different SSH user than root
 
 The default SSH user is root, but you can change it using `ssh/user`:
@@ -193,13 +226,13 @@ ssh:
   user: app
 ```
 
-If you are using non-root user, you need to bootstrap your servers manually, before using them with MRSK. On Ubuntu, you'd do:
+If you are using non-root user (`app` as above example), you need to bootstrap your servers manually, before using them with MRSK. On Ubuntu, you'd do:
 
 ```bash
 sudo apt update
 sudo apt upgrade -y
 sudo apt install -y docker.io curl git
-sudo usermod -a -G docker ubuntu
+sudo usermod -a -G docker app
 ```
 
 ### Using a proxy SSH host
@@ -225,6 +258,15 @@ ssh:
   proxy_command: aws ssm start-session --target %h --document-name AWS-StartSSHSession --parameters 'portNumber=%p' --region=us-east-1 ## ssh via aws ssm
 ```
 
+### Configuring the SSH log level
+
+```yaml
+ssh:
+  log_level: debug
+```
+
+Valid levels are `debug`, `info`, `warn`, `error` and `fatal` (default).
+
 ### Using env variables
 
 You can inject env variables into the app containers using `env`:
@@ -308,7 +350,7 @@ You can specialize the default Traefik rules by setting labels on the containers
 labels:
   traefik.http.routers.hey-web.rule: Host(`app.hey.com`)
 ```
-Traefik rules are in the "service-role-destination" format. The default role will be `web` if no rule is specified. If the destination is not specified, it is not included. To give an example, the above rule would become "traefik.http.routers.hey-web.rule" if it was for the "staging" destination.
+Traefik rules are in the "service-role-destination" format. The default role will be `web` if no rule is specified. If the destination is not specified, it is not included. To give an example, the above rule would become "traefik.http.routers.hey-web-staging.rule" if it was for the "staging" destination.
 
 Note: The backticks are needed to ensure the rule is passed in correctly and not treated as command substitution by Bash!
 
@@ -331,6 +373,21 @@ servers:
       my-label: "50"
 ```
 
+### Using shell expansion
+
+You can use shell expansion to interpolate values from the host machine into labels and env variables with the `${}` syntax.
+Anything within the curly braces will be executed on the host machine and the result will be interpolated into the label or env variable.
+
+```yaml
+labels:
+  host-machine: "${cat /etc/hostname}"
+
+env:
+  HOST_DEPLOYMENT_DIR: "${PWD}"
+```
+
+Note: Any other occurrence of `$` will be escaped to prevent unwanted shell expansion!
+
 ### Using container options
 
 You can specialize the options used to start containers using the `options` definitions:
@@ -352,6 +409,16 @@ servers:
 
 That'll start the job containers with `docker run ... --cap-add --cpu-count 4 ...`.
 
+### Setting a minimum version
+
+You can set the minimum MRSK version with:
+
+```yaml
+minimum_version: 0.13.3
+```
+
+Note: versions <= 0.13.2 will ignore this setting.
+
 ### Configuring logging
 
 You can configure the logging driver and options passed to Docker using `logging`:
@@ -435,6 +502,37 @@ builder:
   context: ".."
 ```
 
+### Using multistage builder cache
+
+Docker multistage build cache can singlehandedly speed up your builds by a lot. Currently MRSK only supports using the GHA cache or the Registry cache:
+
+```yaml
+# Using GHA cache
+builder:
+  cache:
+    type: gha
+
+# Using Registry cache
+builder:
+  cache:
+    type: registry
+
+# Using Registry cache with different cache image
+builder:
+  cache:
+    type: registry
+    # default image name is <image>-build-cache
+    image: application-cache-image
+
+# Using Registry cache with additinonal cache-to options
+builder:
+  cache:
+    type: registry
+    options: mode=max,image-manifest=true,oci-mediatypes=true
+```
+
+For further insights into build cache optimization, check out documentation on Docker's official website: https://docs.docker.com/build/cache/.
+
 ### Using build secrets for new images
 
 Some images need a secret passed in during build time, like a GITHUB_TOKEN, to give access to private gem repositories. This can be done by having the secret in ENV, then referencing it in the builder configuration:
@@ -507,25 +605,25 @@ traefik:
   options:
     publish:
       - 8080:8080
-    volumes:
+    volume:
       - /tmp/example.json:/tmp/example.json
     memory: 512m
 ```
 
 This starts the Traefik container with `--volume /tmp/example.json:/tmp/example.json --publish 8080:8080 --memory 512m` arguments to `docker run`.
 
-### Traefik container lables
+### Traefik container labels
 
 Add labels to Traefik Docker container.
 
 ```yaml
 traefik:
-  lables:
+  labels:
-    - traefik.enable: true
+    traefik.enable: true
-    - traefik.http.routers.dashboard.rule: Host(`traefik.example.com`) && (PathPrefix(`/api`) || PathPrefix(`/dashboard`))
+    traefik.http.routers.dashboard.rule: Host(`traefik.example.com`) && (PathPrefix(`/api`) || PathPrefix(`/dashboard`))
-    - traefik.http.routers.dashboard.service: api@internal
+    traefik.http.routers.dashboard.service: api@internal
-    - traefik.http.routers.dashboard.middlewares: auth
+    traefik.http.routers.dashboard.middlewares: auth
-    - traefik.http.middlewares.auth.basicauth.users: test:$2y$05$H2o72tMaO.TwY1wNQUV1K.fhjRgLHRDWohFvUZOJHBEtUXNKrqUKi # test:password
+    traefik.http.middlewares.auth.basicauth.users: test:$2y$05$H2o72tMaO.TwY1wNQUV1K.fhjRgLHRDWohFvUZOJHBEtUXNKrqUKi # test:password
 ```
 
 This labels Traefik container with `--label traefik.http.routers.dashboard.middlewares=\"auth\"` and so on.
@@ -553,6 +651,16 @@ traefik:
     entrypoints.otherentrypoint.address: ':9000'
 ```
 
+### Rebooting Traefik
+
+If you make changes to Traefik args or labels, you'll need to reboot with:
+
+`mrsk traefik reboot`
+
+In production, reboot the Traefik containers one by one with a slower but safer approach, using a rolling reboot:
+
+`mrsk traefik reboot --rolling`
+
 ### Configuring build args for new images
 
 Build arguments that aren't secret can also be configured:
@@ -640,42 +748,45 @@ servers:
 
 This assumes the Cron settings are stored in `config/crontab`.
 
-### Using audit broadcasts
+### Healthcheck
 
-If you'd like to broadcast audits of deploys, rollbacks, etc to a chatroom or elsewhere, you can configure the `audit_broadcast_cmd` setting with the path to a bin file that will be passed the audit line as the first argument:
+MRSK uses Docker healthchecks to check the health of your application during deployment. Traefik uses this same healthcheck status to determine when a container is ready to receive traffic.
 
-```yaml
+The healthcheck defaults to testing the HTTP response to the path `/up` on port 3000, up to 7 times. You can tailor this behaviour with the `healthcheck` setting:
-audit_broadcast_cmd:
-  bin/audit_broadcast
-```
-
-The broadcast command could look something like:
-
-```bash
-#!/usr/bin/env bash
-curl -q -d content="[My App] ${1}" https://3.basecamp.com/XXXXX/integrations/XXXXX/buckets/XXXXX/chats/XXXXX/lines
-```
-
-That'll post a line like follows to a preconfigured chatbot in Basecamp:
-
-```
-[My App] [dhh] Rolled back to version d264c4e92470ad1bd18590f04466787262f605de
-```
-
-### Custom healthcheck
-
-MRSK defaults to checking the health of your application again `/up` on port 3000 up to 7 times. You can tailor the behaviour with the `healthcheck` setting:
 
 ```yaml
 healthcheck:
   path: /healthz
   port: 4000
   max_attempts: 7
+  interval: 20s
 ```
 
 This will ensure your application is configured with a traefik label for the healthcheck against `/healthz` and that the pre-deploy healthcheck that MRSK performs is done against the same path on port 4000.
 
-The healthcheck also allows for an optional `max_attempts` setting, which will attempt the healthcheck up to the specified number of times before failing the deploy. This is useful for applications that take a while to start up. The default is 7.
+You can also specify a custom healthcheck command, which is useful for non-HTTP services:
+
+```yaml
+healthcheck:
+  cmd: /bin/check_health
+```
+
+The top-level healthcheck configuration applies to all services that use
+Traefik, by default. You can also specialize the configuration at the role
+level:
+
+```yaml
+servers:
+  job:
+    hosts: ...
+    cmd: bin/jobs
+    healthcheck:
+      cmd: bin/check
+```
+
+The healthcheck allows for an optional `max_attempts` setting, which will attempt the healthcheck up to the specified number of times before failing the deploy. This is useful for applications that take a while to start up. The default is 7.
+
+Note: The HTTP health checks assume that the `curl` command is available inside the container. If that's not the case, use the healthcheck's `cmd` option to specify an alternative check that the container supports.
 
 ## Commands
 
@@ -794,7 +905,7 @@ If you wish to remove the entire application, including Traefik, containers, ima
 
 ## Locking
 
-Commands that are unsafe to run concurrently will take a deploy lock while they run. The lock is the `mrsk_lock` directory on the primary server.
+Commands that are unsafe to run concurrently will take a deploy lock while they run. The lock is the `mrsk_lock-<service>` directory on the primary server.
 
 You can check the lock status with:
 
@@ -809,13 +920,85 @@ Message: Automatic deploy lock
 You can also manually acquire and release the lock
 
 ```
-mrsk lock acquire -m "Doing maintanence"
+mrsk lock acquire -m "Doing maintenance"
 ```
 
 ```
 mrsk lock release
 ```
 
+## Rolling deployments
+
+When deploying to large numbers of hosts, you might prefer not to restart your services on every host at the same time.
+
+MRSK's default is to boot new containers on all hosts in parallel. But you can control this by configuring `boot/limit` and `boot/wait` as options:
+
+```yaml
+service: myservice
+
+boot:
+  limit: 10 # Can also specify as a percentage of total hosts, such as "25%"
+  wait: 2
+```
+
+When `limit` is specified, containers will be booted on, at most, `limit` hosts at once. MRSK will pause for `wait` seconds between batches.
+
+These settings only apply when booting containers (using `mrsk deploy`, or `mrsk app boot`). For other commands, MRSK continues to run commands in parallel across all hosts.
+
+## Hooks
+
+You can run custom scripts at specific points with hooks.
+
+Hooks should be stored in the .mrsk/hooks folder. Running mrsk init will build that folder and add some sample scripts.
+
+You can change their location by setting `hooks_path` in the configuration file.
+
+If the script returns a non-zero exit code the command will be aborted.
+
+`MRSK_*` environment variables are available to the hooks command for
+fine-grained audit reporting, e.g. for triggering deployment reports or
+firing a JSON webhook. These variables include:
+- `MRSK_RECORDED_AT` - UTC timestamp in ISO 8601 format, e.g. `2023-04-14T17:07:31Z`
+- `MRSK_PERFORMER` - the local user performing the command (from `whoami`)
+- `MRSK_SERVICE_VERSION` - an abbreviated service and version for use in messages, e.g. app@150b24f
+- `MRSK_VERSION` - the full version being deployed
+- `MRSK_HOSTS` - a comma-separated list of the hosts targeted by the command
+- `MRSK_COMMAND` - The command we are running
+- `MRSK_SUBCOMMAND` - optional: The subcommand we are running
+- `MRSK_DESTINATION` - optional: destination, e.g. "staging"
+- `MRSK_ROLE` - optional: role targeted, e.g. "web"
+
+There are four hooks:
+
+1. pre-connect
+Called before taking the deploy lock. For checks that need to run before connecting to remote hosts - e.g. DNS warming.
+
+2. pre-build
+Used for pre-build checks - e.g. there are no uncommitted changes or that CI has passed.
+
+3. pre-deploy
+For final checks before deploying, e.g. checking CI completed
+
+3. post-deploy - run after a deploy, redeploy or rollback.
+This hook is also passed a `MRSK_RUNTIME` env variable set to the total seconds the deploy took.
+
+This could be used to broadcast a deployment message, or register the new version with an APM.
+
+The command could look something like:
+
+```bash
+#!/usr/bin/env bash
+curl -q -d content="[My App] ${MRSK_PERFORMER} Rolled back to version ${MRSK_VERSION}" https://3.basecamp.com/XXXXX/integrations/XXXXX/buckets/XXXXX/chats/XXXXX/lines
+```
+
+That'll post a line like the following to a preconfigured chatbot in Basecamp:
+
+```
+[My App] [dhh] Rolled back to version d264c4e92470ad1bd18590f04466787262f605de
+```
+
+Set `--skip_hooks` to avoid running the hooks.
+
 ## Stage of development
 
 This is beta software. Commands may still move around. But we're live in production at [37signals](https://37signals.com).

bin/mrsk

@@ -8,7 +8,7 @@ require "mrsk"
 begin
   Mrsk::Cli::Main.start(ARGV)
 rescue SSHKit::Runner::ExecuteError => e
-  puts "  \e[31mERROR (#{e.cause.class}): #{e.cause.message}\e[0m"
+  puts "  \e[31mERROR (#{e.cause.class}): #{e.message}\e[0m"
   puts e.cause.backtrace if ENV["VERBOSE"]
   exit 1
 rescue => e

lib/mrsk/cli.rb

@@ -1,5 +1,6 @@
 module Mrsk::Cli
   class LockError < StandardError; end
+  class HookError < StandardError; end
 end
 
 # SSHKit uses instance eval, so we need a global const for ergonomics

lib/mrsk/cli/accessory.rb

@@ -1,7 +1,7 @@
 class Mrsk::Cli::Accessory < Mrsk::Cli::Base
   desc "boot [NAME]", "Boot new accessory service on host (use NAME=all to boot all accessories)"
-  def boot(name)
+  def boot(name, login: true)
-    with_lock do
+    mutating do
       if name == "all"
         MRSK.accessory_names.each { |accessory_name| boot(accessory_name) }
       else
@@ -10,12 +10,10 @@ class Mrsk::Cli::Accessory < Mrsk::Cli::Base
           upload(name)
 
           on(accessory.hosts) do
-            execute *MRSK.registry.login
+            execute *MRSK.registry.login if login
             execute *MRSK.auditor.record("Booted #{name} accessory"), verbosity: :debug
             execute *accessory.run
           end
-
-          audit_broadcast "Booted accessory #{name}" unless options[:skip_broadcast]
         end
       end
     end
@@ -23,7 +21,7 @@ class Mrsk::Cli::Accessory < Mrsk::Cli::Base
 
   desc "upload [NAME]", "Upload accessory files to host", hide: true
   def upload(name)
-    with_lock do
+    mutating do
       with_accessory(name) do |accessory|
         on(accessory.hosts) do
           accessory.files.each do |(local, remote)|
@@ -40,7 +38,7 @@ class Mrsk::Cli::Accessory < Mrsk::Cli::Base
 
   desc "directories [NAME]", "Create accessory directories on host", hide: true
   def directories(name)
-    with_lock do
+    mutating do
       with_accessory(name) do |accessory|
         on(accessory.hosts) do
           accessory.directories.keys.each do |host_path|
@@ -53,18 +51,22 @@ class Mrsk::Cli::Accessory < Mrsk::Cli::Base
 
   desc "reboot [NAME]", "Reboot existing accessory on host (stop container, remove container, start new container)"
   def reboot(name)
-    with_lock do
+    mutating do
       with_accessory(name) do |accessory|
+        on(accessory.hosts) do
+          execute *MRSK.registry.login
+        end
+
         stop(name)
         remove_container(name)
-        boot(name)
+        boot(name, login: false)
       end
     end
   end
 
   desc "start [NAME]", "Start existing accessory container on host"
   def start(name)
-    with_lock do
+    mutating do
       with_accessory(name) do |accessory|
         on(accessory.hosts) do
           execute *MRSK.auditor.record("Started #{name} accessory"), verbosity: :debug
@@ -76,7 +78,7 @@ class Mrsk::Cli::Accessory < Mrsk::Cli::Base
 
   desc "stop [NAME]", "Stop existing accessory container on host"
   def stop(name)
-    with_lock do
+    mutating do
       with_accessory(name) do |accessory|
         on(accessory.hosts) do
           execute *MRSK.auditor.record("Stopped #{name} accessory"), verbosity: :debug
@@ -88,7 +90,7 @@ class Mrsk::Cli::Accessory < Mrsk::Cli::Base
 
   desc "restart [NAME]", "Restart existing accessory container on host"
   def restart(name)
-    with_lock do
+    mutating do
       with_accessory(name) do
         stop(name)
         start(name)
@@ -167,7 +169,7 @@ class Mrsk::Cli::Accessory < Mrsk::Cli::Base
   desc "remove [NAME]", "Remove accessory container, image and data directory from host (use NAME=all to remove all accessories)"
   option :confirmed, aliases: "-y", type: :boolean, default: false, desc: "Proceed without confirmation question"
   def remove(name)
-    with_lock do
+    mutating do
       if name == "all"
         MRSK.accessory_names.each { |accessory_name| remove(accessory_name) }
       else
@@ -185,7 +187,7 @@ class Mrsk::Cli::Accessory < Mrsk::Cli::Base
 
   desc "remove_container [NAME]", "Remove accessory container from host", hide: true
   def remove_container(name)
-    with_lock do
+    mutating do
       with_accessory(name) do |accessory|
         on(accessory.hosts) do
           execute *MRSK.auditor.record("Remove #{name} accessory container"), verbosity: :debug
@@ -197,7 +199,7 @@ class Mrsk::Cli::Accessory < Mrsk::Cli::Base
 
   desc "remove_image [NAME]", "Remove accessory image from host", hide: true
   def remove_image(name)
-    with_lock do
+    mutating do
       with_accessory(name) do |accessory|
         on(accessory.hosts) do
           execute *MRSK.auditor.record("Removed #{name} accessory image"), verbosity: :debug
@@ -209,7 +211,7 @@ class Mrsk::Cli::Accessory < Mrsk::Cli::Base
 
   desc "remove_service_directory [NAME]", "Remove accessory directory used for uploaded files and data directories from host", hide: true
   def remove_service_directory(name)
-    with_lock do
+    mutating do
       with_accessory(name) do |accessory|
         on(accessory.hosts) do
           execute *accessory.remove_service_directory

lib/mrsk/cli/app.rb

@@ -1,31 +1,39 @@
 class Mrsk::Cli::App < Mrsk::Cli::Base
   desc "boot", "Boot app on servers (or reboot app if already running)"
   def boot
-    with_lock do
+    mutating do
-      say "Get most recent version available as an image...", :magenta unless options[:version]
+      hold_lock_on_error do
-      using_version(version_or_latest) do |version|
+        say "Get most recent version available as an image...", :magenta unless options[:version]
-        say "Start container with version #{version} using a #{MRSK.config.readiness_delay}s readiness delay (or reboot if already running)...", :magenta
+        using_version(version_or_latest) do |version|
+          say "Start container with version #{version} using a #{MRSK.config.readiness_delay}s readiness delay (or reboot if already running)...", :magenta
 
-        cli = self
+          on(MRSK.hosts) do
+            execute *MRSK.auditor.record("Tagging #{MRSK.config.absolute_image} as the latest image"), verbosity: :debug
+            execute *MRSK.app.tag_current_as_latest
+          end
 
-        on(MRSK.hosts) do |host|
+          on(MRSK.hosts, **MRSK.boot_strategy) do |host|
-          roles = MRSK.roles_on(host)
+            roles = MRSK.roles_on(host)
 
-          roles.each do |role|
+            roles.each do |role|
-            execute *MRSK.auditor(role: role).record("Booted app version #{version}"), verbosity: :debug
+              app = MRSK.app(role: role)
+              auditor = MRSK.auditor(role: role)
 
-            begin
+              if capture_with_info(*app.container_id_for_version(version, only_running: true), raise_on_non_zero_exit: false).present?
-              if capture_with_info(*MRSK.app(role: role).container_id_for_version(version)).present?
+                tmp_version = "#{version}_replaced_#{SecureRandom.hex(8)}"
-                tmp_version = "#{version}_#{SecureRandom.hex(8)}"
                 info "Renaming container #{version} to #{tmp_version} as already deployed on #{host}"
-                execute *MRSK.auditor(role: role).record("Renaming container #{version} to #{tmp_version}"), verbosity: :debug
+                execute *auditor.record("Renaming container #{version} to #{tmp_version}"), verbosity: :debug
-                execute *MRSK.app(role: role).rename_container(version: version, new_version: tmp_version)
+                execute *app.rename_container(version: version, new_version: tmp_version)
               end
 
-              old_version = capture_with_info(*MRSK.app(role: role).current_running_version).strip
+              execute *auditor.record("Booted app version #{version}"), verbosity: :debug
-              execute *MRSK.app(role: role).run
+
-              sleep MRSK.config.readiness_delay
+              old_version = capture_with_info(*app.current_running_version, raise_on_non_zero_exit: false).strip
-              execute *MRSK.app(role: role).stop(version: old_version), raise_on_non_zero_exit: false if old_version.present?
+              execute *app.start_or_run(hostname: "#{host}-#{SecureRandom.hex(6)}")
+
+              Mrsk::Utils::HealthcheckPoller.wait_for_healthy(pause_after_ready: true) { capture_with_info(*app.status(version: version)) }
+
+              execute *app.stop(version: old_version), raise_on_non_zero_exit: false if old_version.present?
             end
           end
         end
@@ -35,7 +43,7 @@ class Mrsk::Cli::App < Mrsk::Cli::Base
 
   desc "start", "Start existing app container on servers"
   def start
-    with_lock do
+    mutating do
       on(MRSK.hosts) do |host|
         roles = MRSK.roles_on(host)
 
@@ -49,12 +57,12 @@ class Mrsk::Cli::App < Mrsk::Cli::Base
 
   desc "stop", "Stop app container on servers"
   def stop
-    with_lock do
+    mutating do
       on(MRSK.hosts) do |host|
         roles = MRSK.roles_on(host)
 
         roles.each do |role|
-          execute *MRSK.auditor(role: role).record("Stopped app"), verbosity: :debug
+          execute *MRSK.auditor.record("Stopped app", role: role), verbosity: :debug
           execute *MRSK.app(role: role).stop, raise_on_non_zero_exit: false
         end
       end
@@ -101,7 +109,7 @@ class Mrsk::Cli::App < Mrsk::Cli::Base
           roles = MRSK.roles_on(host)
 
           roles.each do |role|
-            execute *MRSK.auditor(role: role).record("Executed cmd '#{cmd}' on app version #{version}"), verbosity: :debug
+            execute *MRSK.auditor.record("Executed cmd '#{cmd}' on app version #{version}", role: role), verbosity: :debug
             puts_by_host host, capture_with_info(*MRSK.app(role: role).execute_in_existing_container(cmd))
           end
         end
@@ -124,6 +132,31 @@ class Mrsk::Cli::App < Mrsk::Cli::Base
     on(MRSK.hosts) { |host| puts_by_host host, capture_with_info(*MRSK.app.list_containers) }
   end
 
+  desc "stale_containers", "Detect app stale containers"
+  option :stop, aliases: "-s", type: :boolean, default: false, desc: "Stop the stale containers found"
+  def stale_containers
+    mutating do
+      stop = options[:stop]
+
+      cli = self
+
+      on(MRSK.hosts) do |host|
+        roles = MRSK.roles_on(host)
+
+        roles.each do |role|
+          cli.send(:stale_versions, host: host, role: role).each do |version|
+            if stop
+              puts_by_host host, "Stopping stale container for role #{role} with version #{version}"
+              execute *MRSK.app(role: role).stop(version: version), raise_on_non_zero_exit: false
+            else
+              puts_by_host host,  "Detected stale container for role #{role} with version #{version} (use `mrsk app stale_containers --stop` to stop)"
+            end
+          end
+        end
+      end
+    end
+  end
+
   desc "images", "Show app images on servers"
   def images
     on(MRSK.hosts) { |host| puts_by_host host, capture_with_info(*MRSK.app.list_images) }
@@ -169,7 +202,7 @@ class Mrsk::Cli::App < Mrsk::Cli::Base
 
   desc "remove", "Remove app containers and images from servers"
   def remove
-    with_lock do
+    mutating do
       stop
       remove_containers
       remove_images
@@ -178,12 +211,12 @@ class Mrsk::Cli::App < Mrsk::Cli::Base
 
   desc "remove_container [VERSION]", "Remove app container with given version from servers", hide: true
   def remove_container(version)
-    with_lock do
+    mutating do
       on(MRSK.hosts) do |host|
         roles = MRSK.roles_on(host)
 
         roles.each do |role|
-          execute *MRSK.auditor(role: role).record("Removed app container with version #{version}"), verbosity: :debug
+          execute *MRSK.auditor.record("Removed app container with version #{version}", role: role), verbosity: :debug
           execute *MRSK.app(role: role).remove_container(version: version)
         end
       end
@@ -192,12 +225,12 @@ class Mrsk::Cli::App < Mrsk::Cli::Base
 
   desc "remove_containers", "Remove all app containers from servers", hide: true
   def remove_containers
-    with_lock do
+    mutating do
       on(MRSK.hosts) do |host|
         roles = MRSK.roles_on(host)
 
         roles.each do |role|
-          execute *MRSK.auditor(role: role).record("Removed all app containers"), verbosity: :debug
+          execute *MRSK.auditor.record("Removed all app containers", role: role), verbosity: :debug
           execute *MRSK.app(role: role).remove_containers
         end
       end
@@ -206,7 +239,7 @@ class Mrsk::Cli::App < Mrsk::Cli::Base
 
   desc "remove_images", "Remove all app images from servers", hide: true
   def remove_images
-    with_lock do
+    mutating do
       on(MRSK.hosts) do
         execute *MRSK.auditor.record("Removed all app images"), verbosity: :debug
         execute *MRSK.app.remove_images
@@ -240,6 +273,17 @@ class Mrsk::Cli::App < Mrsk::Cli::Base
       version.presence
     end
 
+    def stale_versions(host:, role:)
+      versions = nil
+      on(host) do
+        versions = \
+          capture_with_info(*MRSK.app(role: role).list_versions, raise_on_non_zero_exit: false)
+          .split("\n")
+          .drop(1)
+      end
+      versions
+    end
+
     def version_or_latest
       options[:version] || "latest"
     end

lib/mrsk/cli/base.rb

@@ -20,7 +20,7 @@ module Mrsk::Cli
     class_option :config_file, aliases: "-c", default: "config/deploy.yml", desc: "Path to config file"
     class_option :destination, aliases: "-d", desc: "Specify destination to be used for config file (staging -> deploy.staging.yml)"
 
-    class_option :skip_broadcast, aliases: "-B", type: :boolean, default: false, desc: "Skip audit broadcasts"
+    class_option :skip_hooks, aliases: "-H", type: :boolean, default: false, desc: "Don't run hooks"
 
     def initialize(*)
       super
@@ -72,42 +72,100 @@ module Mrsk::Cli
         puts "  Finished all in #{sprintf("%.1f seconds", runtime)}"
       end
 
-      def audit_broadcast(line)
+      def mutating
-        run_locally { execute *MRSK.auditor.broadcast(line), verbosity: :debug }
+        return yield if MRSK.holding_lock?
-      end
+
+        MRSK.config.ensure_env_available
+
+        run_hook "pre-connect"
 
-      def with_lock
         acquire_lock
 
-        yield
+        begin
+          yield
+        rescue
+          if MRSK.hold_lock_on_error?
+            error "  \e[31mDeploy lock was not released\e[0m"
+          else
+            release_lock
+          end
+
+          raise
+        end
 
         release_lock
-      rescue
-        error "  \e[31mDeploy lock was not released\e[0m" if MRSK.lock_count > 0
-        raise
       end
 
       def acquire_lock
-        if MRSK.lock_count == 0
+        raise_if_locked do
-          say "Acquiring the deploy lock"
+          say "Acquiring the deploy lock...", :magenta
-          on(MRSK.primary_host) { execute *MRSK.lock.acquire("Automatic deploy lock", MRSK.config.version) }
+          on(MRSK.primary_host) { execute *MRSK.lock.acquire("Automatic deploy lock", MRSK.config.version), verbosity: :debug }
         end
-        MRSK.lock_count += 1
+
+        MRSK.holding_lock = true
+      end
+
+      def release_lock
+        say "Releasing the deploy lock...", :magenta
+        on(MRSK.primary_host) { execute *MRSK.lock.release, verbosity: :debug }
+
+        MRSK.holding_lock = false
+      end
+
+      def raise_if_locked
+        yield
       rescue SSHKit::Runner::ExecuteError => e
         if e.message =~ /cannot create directory/
-          invoke "mrsk:cli:lock:status", []
+          on(MRSK.primary_host) { puts capture_with_debug(*MRSK.lock.status) }
           raise LockError, "Deploy lock found"
         else
           raise e
         end
       end
 
-      def release_lock
+      def hold_lock_on_error
-        MRSK.lock_count -= 1
+        if MRSK.hold_lock_on_error?
-        if MRSK.lock_count == 0
+          yield
-          say "Releasing the deploy lock"
+        else
-          on(MRSK.primary_host) { execute *MRSK.lock.release }
+          MRSK.hold_lock_on_error = true
+          yield
+          MRSK.hold_lock_on_error = false
         end
       end
-  end
+
+      def run_hook(hook, **extra_details)
+        if !options[:skip_hooks] && MRSK.hook.hook_exists?(hook)
+          details = { hosts: MRSK.hosts.join(","), command: command, subcommand: subcommand }
+
+          say "Running the #{hook} hook...", :magenta
+          run_locally do
+            MRSK.with_verbosity(:debug) { execute *MRSK.hook.run(hook, **details, **extra_details) }
+          rescue SSHKit::Command::Failed
+            raise HookError.new("Hook `#{hook}` failed")
+          end
+        end
+      end
+
+      def command
+        @mrsk_command ||= begin
+          invocation_class, invocation_commands = *first_invocation
+          if invocation_class == Mrsk::Cli::Main
+            invocation_commands[0]
+          else
+            Mrsk::Cli::Main.subcommand_classes.find { |command, clazz| clazz == invocation_class }[0]
+          end
+        end
+      end
+
+      def subcommand
+        @mrsk_subcommand ||= begin
+          invocation_class, invocation_commands = *first_invocation
+          invocation_commands[0] if invocation_class != Mrsk::Cli::Main
+        end
+      end
+
+      def first_invocation
+        instance_variable_get("@_invocations").first
+      end
+    end
 end

lib/mrsk/cli/build.rb

@@ -1,7 +1,9 @@
 class Mrsk::Cli::Build < Mrsk::Cli::Base
+  class BuildError < StandardError; end
+
   desc "deliver", "Build app and push app image to registry then pull image on servers"
   def deliver
-    with_lock do
+    mutating do
       push
       pull
     end
@@ -9,9 +11,12 @@ class Mrsk::Cli::Build < Mrsk::Cli::Base
 
   desc "push", "Build and push app image to registry"
   def push
-    with_lock do
+    mutating do
       cli = self
 
+      verify_local_dependencies
+      run_hook "pre-build"
+
       run_locally do
         begin
           MRSK.with_verbosity(:debug) { execute *MRSK.builder.push }
@@ -32,7 +37,7 @@ class Mrsk::Cli::Build < Mrsk::Cli::Base
 
   desc "pull", "Pull app image from registry onto servers"
   def pull
-    with_lock do
+    mutating do
       on(MRSK.hosts) do
         execute *MRSK.auditor.record("Pulled image with version #{MRSK.config.version}"), verbosity: :debug
         execute *MRSK.builder.clean, raise_on_non_zero_exit: false
@@ -43,7 +48,7 @@ class Mrsk::Cli::Build < Mrsk::Cli::Base
 
   desc "create", "Create a build setup"
   def create
-    with_lock do
+    mutating do
       run_locally do
         begin
           debug "Using builder: #{MRSK.builder.name}"
@@ -62,7 +67,7 @@ class Mrsk::Cli::Build < Mrsk::Cli::Base
 
   desc "remove", "Remove build setup"
   def remove
-    with_lock do
+    mutating do
       run_locally do
         debug "Using builder: #{MRSK.builder.name}"
         execute *MRSK.builder.remove
@@ -77,4 +82,19 @@ class Mrsk::Cli::Build < Mrsk::Cli::Base
       puts capture(*MRSK.builder.info)
     end
   end
+
+  private
+    def verify_local_dependencies
+      run_locally do
+        begin
+          execute *MRSK.builder.ensure_local_dependencies_installed
+        rescue SSHKit::Command::Failed => e
+          build_error = e.message =~ /command not found/ ?
+            "Docker is not installed locally" :
+            "Docker buildx plugin is not installed locally"
+
+          raise BuildError, build_error
+        end
+      end
+    end
 end

lib/mrsk/cli/healthcheck.rb

@@ -1,7 +1,4 @@
 class Mrsk::Cli::Healthcheck < Mrsk::Cli::Base
-
-  class HealthcheckError < StandardError; end
-
   default_command :perform
 
   desc "perform", "Health check current app version"
@@ -9,38 +6,11 @@ class Mrsk::Cli::Healthcheck < Mrsk::Cli::Base
     on(MRSK.primary_host) do
       begin
         execute *MRSK.healthcheck.run
-
+        Mrsk::Utils::HealthcheckPoller.wait_for_healthy { capture_with_info(*MRSK.healthcheck.status) }
-        target = "Health check against #{MRSK.config.healthcheck["path"]}"
+      rescue Mrsk::Utils::HealthcheckPoller::HealthcheckError => e
-        attempt = 1
-        max_attempts = MRSK.config.healthcheck["max_attempts"]
-
-        begin
-          status = capture_with_info(*MRSK.healthcheck.curl)
-
-          if status == "200"
-            info "#{target} succeeded with 200 OK!"
-          else
-            raise HealthcheckError, "#{target} failed with status #{status}"
-          end
-        rescue SSHKit::Command::Failed
-          if attempt <= max_attempts
-            info "#{target} failed to respond, retrying in #{attempt}s (attempt #{attempt}/#{max_attempts})..."
-            sleep attempt
-            attempt += 1
-
-            retry
-          else
-            raise
-          end
-        end
-      rescue SSHKit::Command::Failed, HealthcheckError => e
         error capture_with_info(*MRSK.healthcheck.logs)
-
+        error capture_with_pretty_json(*MRSK.healthcheck.container_health_log)
-        if e.message =~ /curl/
+        raise
-          raise SSHKit::Command::Failed, "#{target} failed to return 200 OK!"
-        else
-          raise
-        end
       ensure
         execute *MRSK.healthcheck.stop, raise_on_non_zero_exit: false
         execute *MRSK.healthcheck.remove, raise_on_non_zero_exit: false

lib/mrsk/cli/lock.rb

@@ -2,16 +2,16 @@ class Mrsk::Cli::Lock < Mrsk::Cli::Base
   desc "status", "Report lock status"
   def status
     handle_missing_lock do
-      on(MRSK.primary_host) { puts capture_with_info(*MRSK.lock.status) }
+      on(MRSK.primary_host) { puts capture_with_debug(*MRSK.lock.status) }
     end
   end
 
   desc "acquire", "Acquire the deploy lock"
-  option :message, aliases: "-m", type: :string, desc: "A lock mesasge", required: true
+  option :message, aliases: "-m", type: :string, desc: "A lock message", required: true
   def acquire
     message = options[:message]
-    handle_missing_lock do
+    raise_if_locked do
-      on(MRSK.primary_host) { execute *MRSK.lock.acquire(message, MRSK.config.version) }
+      on(MRSK.primary_host) { execute *MRSK.lock.acquire(message, MRSK.config.version), verbosity: :debug }
       say "Acquired the deploy lock"
     end
   end
@@ -19,7 +19,7 @@ class Mrsk::Cli::Lock < Mrsk::Cli::Base
   desc "release", "Release the deploy lock"
   def release
     handle_missing_lock do
-      on(MRSK.primary_host) { execute *MRSK.lock.release }
+      on(MRSK.primary_host) { execute *MRSK.lock.release, verbosity: :debug }
       say "Released the deploy lock"
     end
   end

lib/mrsk/cli/main.rb

@@ -1,8 +1,8 @@
 class Mrsk::Cli::Main < Mrsk::Cli::Base
   desc "setup", "Setup all accessories and deploy app to servers"
   def setup
-    with_lock do
+    print_runtime do
-      print_runtime do
+      mutating do
         invoke "mrsk:cli:server:bootstrap"
         invoke "mrsk:cli:accessory:boot", [ "all" ]
         deploy
@@ -13,12 +13,9 @@ class Mrsk::Cli::Main < Mrsk::Cli::Base
   desc "deploy", "Deploy app to servers"
   option :skip_push, aliases: "-P", type: :boolean, default: false, desc: "Skip image build and push"
   def deploy
-    with_lock do
+    runtime = print_runtime do
-      invoke_options = deploy_options
+      mutating do
-
+        invoke_options = deploy_options
-      runtime = print_runtime do
-        say "Ensure curl and Docker are installed...", :magenta
-        invoke "mrsk:cli:server:bootstrap", [], invoke_options
 
         say "Log into image registry...", :magenta
         invoke "mrsk:cli:registry:login", [], invoke_options
@@ -31,29 +28,34 @@ class Mrsk::Cli::Main < Mrsk::Cli::Base
           invoke "mrsk:cli:build:deliver", [], invoke_options
         end
 
+        run_hook "pre-deploy"
+
         say "Ensure Traefik is running...", :magenta
         invoke "mrsk:cli:traefik:boot", [], invoke_options
 
         say "Ensure app can pass healthcheck...", :magenta
         invoke "mrsk:cli:healthcheck:perform", [], invoke_options
 
+        say "Detect stale containers...", :magenta
+        invoke "mrsk:cli:app:stale_containers", [], invoke_options
+
         invoke "mrsk:cli:app:boot", [], invoke_options
 
         say "Prune old containers and images...", :magenta
         invoke "mrsk:cli:prune:all", [], invoke_options
       end
-
-      audit_broadcast "Deployed #{service_version} in #{runtime.round} seconds" unless options[:skip_broadcast]
     end
+
+    run_hook "post-deploy", runtime: runtime.round
   end
 
   desc "redeploy", "Deploy app to servers without bootstrapping servers, starting Traefik, pruning, and registry login"
   option :skip_push, aliases: "-P", type: :boolean, default: false, desc: "Skip image build and push"
   def redeploy
-    with_lock do
+    runtime = print_runtime do
-      invoke_options = deploy_options
+      mutating do
+        invoke_options = deploy_options
 
-      runtime = print_runtime do
         if options[:skip_push]
           say "Pull app image...", :magenta
           invoke "mrsk:cli:build:pull", [], invoke_options
@@ -62,49 +64,43 @@ class Mrsk::Cli::Main < Mrsk::Cli::Base
           invoke "mrsk:cli:build:deliver", [], invoke_options
         end
 
+        run_hook "pre-deploy"
+
         say "Ensure app can pass healthcheck...", :magenta
         invoke "mrsk:cli:healthcheck:perform", [], invoke_options
 
+        say "Detect stale containers...", :magenta
+        invoke "mrsk:cli:app:stale_containers", [], invoke_options
+
         invoke "mrsk:cli:app:boot", [], invoke_options
       end
-
-      audit_broadcast "Redeployed #{service_version} in #{runtime.round} seconds" unless options[:skip_broadcast]
     end
+
+    run_hook "post-deploy", runtime: runtime.round
   end
 
   desc "rollback [VERSION]", "Rollback app to VERSION"
   def rollback(version)
-    with_lock do
+    rolled_back = false
-      MRSK.config.version = version
+    runtime = print_runtime do
+      mutating do
+        invoke_options = deploy_options
 
-      if container_available?(version)
+        MRSK.config.version = version
-        say "Start version #{version}, then wait #{MRSK.config.readiness_delay}s for app to boot before stopping the old version...", :magenta
-
-        cli = self
         old_version = nil
 
-        on(MRSK.hosts) do |host|
+        if container_available?(version)
-          roles = MRSK.roles_on(host)
+          run_hook "pre-deploy"
 
-          roles.each do |role|
+          invoke "mrsk:cli:app:boot", [], invoke_options.merge(version: version)
-            app = MRSK.app(role: role)
+          rolled_back = true
-            old_version = capture_with_info(*app.current_running_version).strip.presence
+        else
-
+          say "The app version '#{version}' is not available as a container (use 'mrsk app containers' for available versions)", :red
-            execute *app.start
-
-            if old_version
-              sleep MRSK.config.readiness_delay
-
-              execute *app.stop(version: old_version), raise_on_non_zero_exit: false
-            end
-          end
         end
-
-        audit_broadcast "Rolled back #{service_version(Mrsk::Utils.abbreviate_version(old_version))} to #{service_version}" unless options[:skip_broadcast]
-      else
-        say "The app version '#{version}' is not available as a container (use 'mrsk app containers' for available versions)", :red
       end
     end
+
+    run_hook "post-deploy", runtime: runtime.round if rolled_back
   end
 
   desc "details", "Show details about all containers"
@@ -146,6 +142,14 @@ class Mrsk::Cli::Main < Mrsk::Cli::Base
       puts "Created .env file"
     end
 
+    unless (hooks_dir = Pathname.new(File.expand_path(".mrsk/hooks"))).exist?
+      hooks_dir.mkpath
+      Pathname.new(File.expand_path("templates/sample_hooks", __dir__)).each_child do |sample_hook|
+        FileUtils.cp sample_hook, hooks_dir, preserve: true
+      end
+      puts "Created sample hooks in .mrsk/hooks"
+    end
+
     if options[:bundle]
       if (binstub = Pathname.new(File.expand_path("bin/mrsk"))).exist?
         puts "Binstub already exists in bin/mrsk (remove first to create a new one)"
@@ -176,7 +180,7 @@ class Mrsk::Cli::Main < Mrsk::Cli::Base
   desc "remove", "Remove Traefik, app, accessories, and registry session from servers"
   option :confirmed, aliases: "-y", type: :boolean, default: false, desc: "Proceed without confirmation question"
   def remove
-    with_lock do
+    mutating do
       if options[:confirmed] || ask("This will remove all containers and images. Are you sure?", limited_to: %w( y N ), default: "N") == "y"
         invoke "mrsk:cli:traefik:remove", [], options.without(:confirmed)
         invoke "mrsk:cli:app:remove", [], options.without(:confirmed)
@@ -203,6 +207,9 @@ class Mrsk::Cli::Main < Mrsk::Cli::Base
   desc "healthcheck", "Healthcheck application"
   subcommand "healthcheck", Mrsk::Cli::Healthcheck
 
+  desc "lock", "Manage the deploy lock"
+  subcommand "lock", Mrsk::Cli::Lock
+
   desc "prune", "Prune old application images and containers"
   subcommand "prune", Mrsk::Cli::Prune
 
@@ -215,26 +222,28 @@ class Mrsk::Cli::Main < Mrsk::Cli::Base
   desc "traefik", "Manage Traefik load balancer"
   subcommand "traefik", Mrsk::Cli::Traefik
 
-  desc "lock", "Manage the deploy lock"
-  subcommand "lock", Mrsk::Cli::Lock
-
   private
-    def container_available?(version, host: MRSK.primary_host)
+    def container_available?(version)
-      available = nil
+      begin
-
+        on(MRSK.hosts) do
-      on(host) do
+          MRSK.roles_on(host).each do |role|
-        first_role = MRSK.roles_on(host).first
+            container_id = capture_with_info(*MRSK.app(role: role).container_id_for_version(version))
-        available = capture_with_info(*MRSK.app(role: first_role).container_id_for_version(version)).present?
+            raise "Container not found" unless container_id.present?
+          end
+        end
+      rescue SSHKit::Runner::ExecuteError => e
+        if e.message =~ /Container not found/
+          say "Error looking for container version #{version}: #{e.message}"
+          return false
+        else
+          raise
+        end
       end
 
-      available
+      true
     end
 
     def deploy_options
       { "version" => MRSK.config.version }.merge(options.without("skip_push"))
     end
-
-    def service_version(version = MRSK.config.abbreviated_version)
-      [ MRSK.config.service, version ].compact.join("@")
-    end
 end

lib/mrsk/cli/prune.rb

@@ -1,25 +1,26 @@
 class Mrsk::Cli::Prune < Mrsk::Cli::Base
   desc "all", "Prune unused images and stopped containers"
   def all
-    with_lock do
+    mutating do
       containers
       images
     end
   end
 
-  desc "images", "Prune unused images older than 7 days"
+  desc "images", "Prune dangling images"
   def images
-    with_lock do
+    mutating do
       on(MRSK.hosts) do
         execute *MRSK.auditor.record("Pruned images"), verbosity: :debug
-        execute *MRSK.prune.images
+        execute *MRSK.prune.dangling_images
+        execute *MRSK.prune.tagged_images
       end
     end
   end
 
-  desc "containers", "Prune stopped containers older than 3 days"
+  desc "containers", "Prune all stopped containers, except the last 5"
   def containers
-    with_lock do
+    mutating do
       on(MRSK.hosts) do
         execute *MRSK.auditor.record("Pruned containers"), verbosity: :debug
         execute *MRSK.prune.containers

lib/mrsk/cli/server.rb

@@ -1,17 +1,21 @@
 class Mrsk::Cli::Server < Mrsk::Cli::Base
-  desc "bootstrap", "Ensure curl and Docker are installed on servers"
+  desc "bootstrap", "Set up Docker to run MRSK apps"
   def bootstrap
-    with_lock do
+    missing = []
-      on(MRSK.hosts + MRSK.accessory_hosts) do
-        dependencies_to_install = Array.new.tap do |dependencies|
-          dependencies << "curl" unless execute "which curl", raise_on_non_zero_exit: false
-          dependencies << "docker.io" unless execute "which docker", raise_on_non_zero_exit: false
-        end
 
-        if dependencies_to_install.any?
+    on(MRSK.hosts | MRSK.accessory_hosts) do |host|
-          execute "apt-get update -y && apt-get install #{dependencies_to_install.join(" ")} -y"
+      unless execute(*MRSK.docker.installed?, raise_on_non_zero_exit: false)
+        if execute(*MRSK.docker.superuser?, raise_on_non_zero_exit: false)
+          info "Missing Docker on #{host}. Installing…"
+          execute *MRSK.docker.install
+        else
+          missing << host
         end
       end
     end
+
+    if missing.any?
+      raise "Docker is not installed on #{missing.join(", ")} and can't be automatically installed without having root access and the `curl` command available. Install Docker manually: https://docs.docker.com/engine/install/"
+    end
   end
 end

lib/mrsk/cli/templates/deploy.yml

@@ -25,10 +25,6 @@ registry:
 #   secret:
 #     - RAILS_MASTER_KEY
 
-# Call a broadcast command on deploys.
-# audit_broadcast_cmd:
-#   bin/broadcast_to_bc
-
 # Use a different ssh user than root
 # ssh:
 #   user: app

lib/mrsk/cli/templates/sample_hooks/post-deploy.sample

@@ -0,0 +1,14 @@
+#!/bin/sh
+
+# A sample post-deploy hook
+#
+# These environment variables are available:
+# MRSK_RECORDED_AT
+# MRSK_PERFORMER
+# MRSK_VERSION
+# MRSK_HOSTS
+# MRSK_ROLE (if set)
+# MRSK_DESTINATION (if set)
+# MRSK_RUNTIME
+
+echo "$MRSK_PERFORMER deployed $MRSK_VERSION to $MRSK_DESTINATION in $MRSK_RUNTIME seconds"

lib/mrsk/cli/templates/sample_hooks/pre-build.sample

@@ -0,0 +1,51 @@
+#!/bin/sh
+
+# A sample pre-build hook
+#
+# Checks:
+# 1. We have a clean checkout
+# 2. A remote is configured
+# 3. The branch has been pushed to the remote
+# 4. The version we are deploying matches the remote
+#
+# These environment variables are available:
+# MRSK_RECORDED_AT
+# MRSK_PERFORMER
+# MRSK_VERSION
+# MRSK_HOSTS
+# MRSK_ROLE (if set)
+# MRSK_DESTINATION (if set)
+
+if [ -n "$(git status --porcelain)" ]; then
+  echo "Git checkout is not clean, aborting..." >&2
+  git status --porcelain >&2
+  exit 1
+fi
+
+first_remote=$(git remote)
+
+if [ -z "$first_remote" ]; then
+  echo "No git remote set, aborting..." >&2
+  exit 1
+fi
+
+current_branch=$(git branch --show-current)
+
+if [ -z "$current_branch" ]; then
+  echo "No git remote set, aborting..." >&2
+  exit 1
+fi
+
+remote_head=$(git ls-remote $first_remote --tags $current_branch | cut -f1)
+
+if [ -z "$remote_head" ]; then
+  echo "Branch not pushed to remote, aborting..." >&2
+  exit 1
+fi
+
+if [ "$MRSK_VERSION" != "$remote_head" ]; then
+  echo "Version ($MRSK_VERSION) does not match remote HEAD ($remote_head), aborting..." >&2
+  exit 1
+fi
+
+exit 0

lib/mrsk/cli/templates/sample_hooks/pre-connect.sample

@@ -0,0 +1,47 @@
+#!/usr/bin/env ruby
+
+# A sample pre-connect check
+#
+# Warms DNS before connecting to hosts in parallel
+#
+# These environment variables are available:
+# MRSK_RECORDED_AT
+# MRSK_PERFORMER
+# MRSK_VERSION
+# MRSK_HOSTS
+# MRSK_ROLE (if set)
+# MRSK_DESTINATION (if set)
+# MRSK_RUNTIME
+
+hosts = ENV["MRSK_HOSTS"].split(",")
+results = nil
+max = 3
+
+elapsed = Benchmark.realtime do
+  results = hosts.map do |host|
+    Thread.new do
+      tries = 1
+
+      begin
+        Socket.getaddrinfo(host, 0, Socket::AF_UNSPEC, Socket::SOCK_STREAM, nil, Socket::AI_CANONNAME)
+      rescue SocketError
+        if tries < max
+          puts "Retrying DNS warmup: #{host}"
+          tries += 1
+          sleep rand
+          retry
+        else
+          puts "DNS warmup failed: #{host}"
+          host
+        end
+      end
+
+      tries
+    end
+  end.map(&:value)
+end
+
+retries = results.sum - hosts.size
+nopes = results.count { |r| r == max }
+
+puts "Prewarmed %d DNS lookups in %.2f sec: %d retries, %d failures" % [ hosts.size, elapsed, retries, nopes ]

lib/mrsk/cli/templates/sample_hooks/pre-deploy.sample

@@ -0,0 +1,109 @@
+#!/usr/bin/env ruby
+
+# A sample pre-deploy hook
+#
+# Checks the Github status of the build, waiting for a pending build to complete for up to 720 seconds.
+#
+# Fails unless the combined status is "success"
+#
+# These environment variables are available:
+# MRSK_RECORDED_AT
+# MRSK_PERFORMER
+# MRSK_VERSION
+# MRSK_HOSTS
+# MRSK_COMMAND
+# MRSK_SUBCOMMAND
+# MRSK_ROLE (if set)
+# MRSK_DESTINATION (if set)
+
+# Only check the build status for production deployments
+if ENV["MRSK_COMMAND"] == "rollback" || ENV["MRSK_DESTINATION"] != "production"
+  exit 0
+end
+
+require "bundler/inline"
+
+# true = install gems so this is fast on repeat invocations
+gemfile(true, quiet: true) do
+  source "https://rubygems.org"
+
+  gem "octokit"
+  gem "faraday-retry"
+end
+
+MAX_ATTEMPTS = 72
+ATTEMPTS_GAP = 10
+
+def exit_with_error(message)
+  $stderr.puts message
+  exit 1
+end
+
+class GithubStatusChecks
+  attr_reader :remote_url, :git_sha, :github_client, :combined_status
+
+  def initialize
+    @remote_url = `git config --get remote.origin.url`.strip.delete_prefix("https://github.com/")
+    @git_sha = `git rev-parse HEAD`.strip
+    @github_client = Octokit::Client.new(access_token: ENV["GITHUB_TOKEN"])
+    refresh!
+  end
+
+  def refresh!
+    @combined_status = github_client.combined_status(remote_url, git_sha)
+  end
+
+  def state
+    combined_status[:state]
+  end
+
+  def first_status_url
+    first_status = combined_status[:statuses].find { |status| status[:state] == state }
+    first_status && first_status[:target_url]
+  end
+
+  def complete_count
+    combined_status[:statuses].count { |status| status[:state] != "pending"}
+  end
+
+  def total_count
+    combined_status[:statuses].count
+  end
+
+  def current_status
+    if total_count > 0
+      "Completed #{complete_count}/#{total_count} checks, see #{first_status_url} ..."
+    else
+      "Build not started..."
+    end
+  end
+end
+
+
+$stdout.sync = true
+
+puts "Checking build status..."
+attempts = 0
+checks = GithubStatusChecks.new
+
+begin
+  loop do
+    case checks.state
+    when "success"
+      puts "Checks passed, see #{checks.first_status_url}"
+      exit 0
+    when "failure"
+      exit_with_error "Checks failed, see #{checks.first_status_url}"
+    when "pending"
+      attempts += 1
+    end
+
+    exit_with_error "Checks are still pending, gave up after #{MAX_ATTEMPTS * ATTEMPTS_GAP} seconds" if attempts == MAX_ATTEMPTS
+
+    puts checks.current_status
+    sleep(ATTEMPTS_GAP)
+    checks.refresh!
+  end
+rescue Octokit::NotFound
+  exit_with_error "Build status could not be found"
+end

lib/mrsk/cli/traefik.rb

@@ -1,7 +1,7 @@
 class Mrsk::Cli::Traefik < Mrsk::Cli::Base
   desc "boot", "Boot Traefik on servers"
   def boot
-    with_lock do
+    mutating do
       on(MRSK.traefik_hosts) do
         execute *MRSK.registry.login
         execute *MRSK.traefik.run, raise_on_non_zero_exit: false
@@ -10,17 +10,22 @@ class Mrsk::Cli::Traefik < Mrsk::Cli::Base
   end
 
   desc "reboot", "Reboot Traefik on servers (stop container, remove container, start new container)"
+  option :rolling, type: :boolean, default: false, desc: "Reboot traefik on hosts in sequence, rather than in parallel"
   def reboot
-    with_lock do
+    mutating do
-      stop
+      on(MRSK.traefik_hosts, in: options[:rolling] ? :sequence : :parallel) do
-      remove_container
+        execute *MRSK.auditor.record("Rebooted traefik"), verbosity: :debug
-      boot
+        execute *MRSK.registry.login
+        execute *MRSK.traefik.stop, raise_on_non_zero_exit: false
+        execute *MRSK.traefik.remove_container
+        execute *MRSK.traefik.run, raise_on_non_zero_exit: false
+      end
     end
   end
 
   desc "start", "Start existing Traefik container on servers"
   def start
-    with_lock do
+    mutating do
       on(MRSK.traefik_hosts) do
         execute *MRSK.auditor.record("Started traefik"), verbosity: :debug
         execute *MRSK.traefik.start, raise_on_non_zero_exit: false
@@ -30,7 +35,7 @@ class Mrsk::Cli::Traefik < Mrsk::Cli::Base
 
   desc "stop", "Stop existing Traefik container on servers"
   def stop
-    with_lock do
+    mutating do
       on(MRSK.traefik_hosts) do
         execute *MRSK.auditor.record("Stopped traefik"), verbosity: :debug
         execute *MRSK.traefik.stop, raise_on_non_zero_exit: false
@@ -40,7 +45,7 @@ class Mrsk::Cli::Traefik < Mrsk::Cli::Base
 
   desc "restart", "Restart existing Traefik container on servers"
   def restart
-    with_lock do
+    mutating do
       stop
       start
     end
@@ -77,7 +82,7 @@ class Mrsk::Cli::Traefik < Mrsk::Cli::Base
 
   desc "remove", "Remove Traefik container and image from servers"
   def remove
-    with_lock do
+    mutating do
       stop
       remove_container
       remove_image
@@ -86,7 +91,7 @@ class Mrsk::Cli::Traefik < Mrsk::Cli::Base
 
   desc "remove_container", "Remove Traefik container from servers", hide: true
   def remove_container
-    with_lock do
+    mutating do
       on(MRSK.traefik_hosts) do
         execute *MRSK.auditor.record("Removed traefik container"), verbosity: :debug
         execute *MRSK.traefik.remove_container
@@ -94,9 +99,9 @@ class Mrsk::Cli::Traefik < Mrsk::Cli::Base
     end
   end
 
-  desc "remove_container", "Remove Traefik image from servers", hide: true
+  desc "remove_image", "Remove Traefik image from servers", hide: true
   def remove_image
-    with_lock do
+    mutating do
       on(MRSK.traefik_hosts) do
         execute *MRSK.auditor.record("Removed traefik image"), verbosity: :debug
         execute *MRSK.traefik.remove_image

lib/mrsk/commander.rb

@@ -2,11 +2,12 @@ require "active_support/core_ext/enumerable"
 require "active_support/core_ext/module/delegation"
 
 class Mrsk::Commander
-  attr_accessor :verbosity, :lock_count
+  attr_accessor :verbosity, :holding_lock, :hold_lock_on_error
 
   def initialize
     self.verbosity = :info
-    self.lock_count = 0
+    self.holding_lock = false
+    self.hold_lock_on_error = false
   end
 
   def config
@@ -35,7 +36,7 @@ class Mrsk::Commander
   end
 
   def primary_host
-    specific_hosts&.first || config.primary_web_host
+    specific_hosts&.first || specific_roles&.first&.primary_host || config.primary_web_host
   end
 
   def roles
@@ -50,6 +51,14 @@ class Mrsk::Commander
     end
   end
 
+  def boot_strategy
+    if config.boot.limit.present?
+      { in: :groups, limit: config.boot.limit, wait: config.boot.wait }
+    else
+      {}
+    end
+  end
+
   def roles_on(host)
     roles.select { |role| role.hosts.include?(host.to_s) }.map(&:name)
   end
@@ -75,18 +84,30 @@ class Mrsk::Commander
     Mrsk::Commands::Accessory.new(config, name: name)
   end
 
-  def auditor(role: nil)
+  def auditor(**details)
-    Mrsk::Commands::Auditor.new(config, role: role)
+    Mrsk::Commands::Auditor.new(config, **details)
   end
 
   def builder
     @builder ||= Mrsk::Commands::Builder.new(config)
   end
 
+  def docker
+    @docker ||= Mrsk::Commands::Docker.new(config)
+  end
+
   def healthcheck
     @healthcheck ||= Mrsk::Commands::Healthcheck.new(config)
   end
 
+  def hook
+    @hook ||= Mrsk::Commands::Hook.new(config)
+  end
+
+  def lock
+    @lock ||= Mrsk::Commands::Lock.new(config)
+  end
+
   def prune
     @prune ||= Mrsk::Commands::Prune.new(config)
   end
@@ -99,10 +120,6 @@ class Mrsk::Commander
     @traefik ||= Mrsk::Commands::Traefik.new(config)
   end
 
-  def lock
-    @lock ||= Mrsk::Commands::Lock.new(config)
-  end
-
   def with_verbosity(level)
     old_level = self.verbosity
 
@@ -115,6 +132,14 @@ class Mrsk::Commander
     SSHKit.config.output_verbosity = old_level
   end
 
+  def holding_lock?
+    self.holding_lock
+  end
+
+  def hold_lock_on_error?
+    self.hold_lock_on_error
+  end
+
   private
     # Lazy setup of SSHKit
     def configure_sshkit_with(config)

lib/mrsk/commands/app.rb

@@ -1,4 +1,6 @@
 class Mrsk::Commands::App < Mrsk::Commands::Base
+  ACTIVE_DOCKER_STATUSES = [ :running, :restarting ]
+
   attr_reader :role
 
   def initialize(config, role: nil)
@@ -6,15 +8,21 @@ class Mrsk::Commands::App < Mrsk::Commands::Base
     @role = role
   end
 
-  def run
+  def start_or_run(hostname: nil)
+    combine start, run(hostname: hostname), by: "||"
+  end
+
+  def run(hostname: nil)
     role = config.role(self.role)
 
     docker :run,
       "--detach",
       "--restart unless-stopped",
       "--name", container_name,
+      *(["--hostname", hostname] if hostname),
       "-e", "MRSK_CONTAINER_NAME=\"#{container_name}\"",
       *role.env_args,
+      *role.health_check_args,
       *config.logging_args,
       *config.volume_args,
       *role.label_args,
@@ -27,9 +35,13 @@ class Mrsk::Commands::App < Mrsk::Commands::Base
     docker :start, container_name
   end
 
+  def status(version:)
+    pipe container_id_for_version(version), xargs(docker(:inspect, "--format", DOCKER_HEALTH_STATUS_FORMAT))
+  end
+
   def stop(version: nil)
     pipe \
-      version ? container_id_for_version(version) : current_container_id,
+      version ? container_id_for_version(version) : current_running_container_id,
       xargs(config.stop_wait_time ? docker(:stop, "-t", config.stop_wait_time) : docker(:stop))
   end
 
@@ -40,7 +52,7 @@ class Mrsk::Commands::App < Mrsk::Commands::Base
 
   def logs(since: nil, lines: nil, grep: nil)
     pipe \
-      current_container_id,
+      current_running_container_id,
       "xargs docker logs#{" --since #{since}" if since}#{" --tail #{lines}" if lines} 2>&1",
       ("grep '#{grep}'" if grep)
   end
@@ -48,7 +60,7 @@ class Mrsk::Commands::App < Mrsk::Commands::Base
   def follow_logs(host:, grep: nil)
     run_over_ssh \
       pipe(
-        current_container_id,
+        current_running_container_id,
         "xargs docker logs --timestamps --tail 10 --follow 2>&1",
         (%(grep "#{grep}") if grep)
       ),
@@ -82,20 +94,23 @@ class Mrsk::Commands::App < Mrsk::Commands::Base
   end
 
 
-  def current_container_id
+  def current_running_container_id
-    docker :ps, "--quiet", *filter_args
+    docker :ps, "--quiet", *filter_args(statuses: ACTIVE_DOCKER_STATUSES), "--latest"
   end
 
-  def container_id_for_version(version)
+  def container_id_for_version(version, only_running: false)
-    container_id_for(container_name: container_name(version))
+    container_id_for(container_name: container_name(version), only_running: only_running)
   end
 
   def current_running_version
-    # FIXME: Find more graceful way to extract the version from "app-version" than using sed and tail!
+    list_versions("--latest", statuses: ACTIVE_DOCKER_STATUSES)
+  end
+
+  def list_versions(*docker_args, statuses: nil)
     pipe \
-      docker(:ps, *filter_args, "--format", '"{{.Names}}"'),
+      docker(:ps, *filter_args(statuses: statuses), *docker_args, "--format", '"{{.Names}}"'),
-      %(sed 's/-/\\n/g'),
+      %(grep -oE "\\-[^-]+$"), # Extract SHA from "service-role-dest-SHA"
-      "tail -n 1"
+      %(cut -c 2-)
   end
 
   def list_containers
@@ -128,20 +143,27 @@ class Mrsk::Commands::App < Mrsk::Commands::Base
     docker :image, :prune, "--all", "--force", *filter_args
   end
 
+  def tag_current_as_latest
+    docker :tag, config.absolute_image, config.latest_image
+  end
+
 
   private
     def container_name(version = nil)
       [ config.service, role, config.destination, version || config.version ].compact.join("-")
     end
 
-    def filter_args
+    def filter_args(statuses: nil)
-      argumentize "--filter", filters
+      argumentize "--filter", filters(statuses: statuses)
     end
 
-    def filters
+    def filters(statuses: nil)
       [ "label=service=#{config.service}" ].tap do |filters|
         filters << "label=destination=#{config.destination}" if config.destination
         filters << "label=role=#{role}" if role
+        statuses&.each do |status|
+          filters << "status=#{status}"
+        end
       end
     end
 end

lib/mrsk/commands/auditor.rb

@@ -1,27 +1,18 @@
-require "active_support/core_ext/time/conversions"
-
 class Mrsk::Commands::Auditor < Mrsk::Commands::Base
-  attr_reader :role
+  attr_reader :details
 
-  def initialize(config, role: nil)
+  def initialize(config, **details)
     super(config)
-    @role = role
+    @details = details
   end
 
   # Runs remotely
-  def record(line)
+  def record(line, **details)
     append \
-      [ :echo, tagged_record_line(line) ],
+      [ :echo, audit_tags(**details).except(:version, :service_version).to_s, line ],
       audit_log_file
   end
 
-  # Runs locally
-  def broadcast(line)
-    if broadcast_cmd = config.audit_broadcast_cmd
-      [ broadcast_cmd, tagged_broadcast_line(line) ]
-    end
-  end
-
   def reveal
     [ :tail, "-n", 50, audit_log_file ]
   end
@@ -31,27 +22,7 @@ class Mrsk::Commands::Auditor < Mrsk::Commands::Base
       [ "mrsk", config.service, config.destination, "audit.log" ].compact.join("-")
     end
 
-    def tagged_record_line(line)
+    def audit_tags(**details)
-      tagged_line recorded_at_tag, performer_tag, role_tag, line
+      tags(**self.details, **details)
-    end
-
-    def tagged_broadcast_line(line)
-      tagged_line performer_tag, role_tag, line
-    end
-
-    def tagged_line(*tags_and_line)
-      "'#{tags_and_line.compact.join(" ")}'"
-    end
-
-    def recorded_at_tag
-      "[#{Time.now.to_fs(:db)}]"
-    end
-
-    def performer_tag
-      "[#{`whoami`.strip}]"
-    end
-
-    def role_tag
-      "[#{role}]" if role
     end
 end

lib/mrsk/commands/base.rb

@@ -2,6 +2,9 @@ module Mrsk::Commands
   class Base
     delegate :sensitive, :argumentize, to: Mrsk::Utils
 
+    DOCKER_HEALTH_STATUS_FORMAT = "'{{if .State.Health}}{{.State.Health.Status}}{{else}}{{.State.Status}}{{end}}'"
+    DOCKER_HEALTH_LOG_FORMAT    = "'{{json .State.Health}}'"
+
     attr_accessor :config
 
     def initialize(config)
@@ -10,13 +13,17 @@ module Mrsk::Commands
 
     def run_over_ssh(*command, host:)
       "ssh".tap do |cmd|
-        cmd << " -J #{config.ssh_proxy.jump_proxies}" if config.ssh_proxy
+        if config.ssh_proxy && config.ssh_proxy.is_a?(Net::SSH::Proxy::Jump)
+          cmd << " -J #{config.ssh_proxy.jump_proxies}"
+        elsif config.ssh_proxy && config.ssh_proxy.is_a?(Net::SSH::Proxy::Command)
+          cmd << " -o ProxyCommand='#{config.ssh_proxy.command_line_template}'"
+        end
         cmd << " -t #{config.ssh_user}@#{host} '#{command.join(" ")}'"
       end
     end
 
-    def container_id_for(container_name:)
+    def container_id_for(container_name:, only_running: false)
-      docker :container, :ls, "--all", "--filter", "name=^#{container_name}$", "--quiet"
+      docker :container, :ls, *("--all" unless only_running), "--filter", "name=^#{container_name}$", "--quiet"
     end
 
     private
@@ -50,5 +57,9 @@ module Mrsk::Commands
       def docker(*args)
         args.compact.unshift :docker
       end
+
+      def tags(**details)
+        Mrsk::Tags.from_config(config, **details)
+      end
   end
 end

lib/mrsk/commands/builder.rb

@@ -2,16 +2,18 @@ class Mrsk::Commands::Builder < Mrsk::Commands::Base
   delegate :create, :remove, :push, :clean, :pull, :info, to: :target
 
   def name
-    target.class.to_s.remove("Mrsk::Commands::Builder::").underscore
+    target.class.to_s.remove("Mrsk::Commands::Builder::").underscore.inquiry
   end
 
   def target
     case
-    when config.builder && config.builder["multiarch"] == false
+    when !config.builder.multiarch? && !config.builder.cached?
       native
-    when config.builder && config.builder["local"] && config.builder["remote"]
+    when !config.builder.multiarch? && config.builder.cached?
+      native_cached
+    when config.builder.local? && config.builder.remote?
       multiarch_remote
-    when config.builder && config.builder["remote"]
+    when config.builder.remote?
       native_remote
     else
       multiarch
@@ -22,6 +24,10 @@ class Mrsk::Commands::Builder < Mrsk::Commands::Base
     @native ||= Mrsk::Commands::Builder::Native.new(config)
   end
 
+  def native_cached
+    @native ||= Mrsk::Commands::Builder::Native::Cached.new(config)
+  end
+
   def native_remote
     @native ||= Mrsk::Commands::Builder::Native::Remote.new(config)
   end
@@ -33,4 +39,24 @@ class Mrsk::Commands::Builder < Mrsk::Commands::Base
   def multiarch_remote
     @multiarch_remote ||= Mrsk::Commands::Builder::Multiarch::Remote.new(config)
   end
+
+
+  def ensure_local_dependencies_installed
+    if name.native?
+      ensure_local_docker_installed
+    else
+      combine \
+        ensure_local_docker_installed,
+        ensure_local_buildx_installed
+    end
+  end
+
+  private
+    def ensure_local_docker_installed
+      docker "--version"
+    end
+
+    def ensure_local_buildx_installed
+      docker :buildx, "version"
+    end
 end

lib/mrsk/commands/builder/base.rb

@@ -1,5 +1,9 @@
+
 class Mrsk::Commands::Builder::Base < Mrsk::Commands::Base
+  class BuilderError < StandardError; end
+
   delegate :argumentize, to: Mrsk::Utils
+  delegate :args, :secrets, :dockerfile, :local_arch, :local_host, :remote_arch, :remote_host, :cache_from, :cache_to, to: :builder_config
 
   def clean
     docker :image, :rm, "--force", config.absolute_image
@@ -7,22 +11,29 @@ class Mrsk::Commands::Builder::Base < Mrsk::Commands::Base
 
   def pull
     docker :pull, config.absolute_image
-    docker :pull, config.latest_image
   end
 
   def build_options
-    [ *build_tags, *build_labels, *build_args, *build_secrets, *build_dockerfile ]
+    [ *build_tags, *build_cache, *build_labels, *build_args, *build_secrets, *build_dockerfile ]
   end
 
   def build_context
-    context
+    config.builder.context
   end
 
+
   private
     def build_tags
       [ "-t", config.absolute_image, "-t", config.latest_image ]
     end
 
+    def build_cache
+      if cache_to && cache_from
+        ["--cache-to", cache_to,
+          "--cache-from", cache_from]
+      end
+    end
+
     def build_labels
       argumentize "--label", { service: config.service }
     end
@@ -36,22 +47,14 @@ class Mrsk::Commands::Builder::Base < Mrsk::Commands::Base
     end
 
     def build_dockerfile
-      argumentize "--file", dockerfile
+      if Pathname.new(File.expand_path(dockerfile)).exist?
+        argumentize "--file", dockerfile
+      else
+        raise BuilderError, "Missing #{dockerfile}"
+      end
     end
 
-    def args
+    def builder_config
-      (config.builder && config.builder["args"]) || {}
+      config.builder
-    end
-
-    def secrets
-      (config.builder && config.builder["secrets"]) || []
-    end
-
-    def dockerfile
-      (config.builder && config.builder["dockerfile"]) || "Dockerfile"
-    end
-
-    def context
-      (config.builder && config.builder["context"]) || "."
     end
 end

lib/mrsk/commands/builder/multiarch/remote.rb

@@ -22,17 +22,17 @@ class Mrsk::Commands::Builder::Multiarch::Remote < Mrsk::Commands::Builder::Mult
     end
 
     def create_local_buildx
-      docker :buildx, :create, "--name", builder_name, builder_name_with_arch(local["arch"]), "--platform", "linux/#{local["arch"]}"
+      docker :buildx, :create, "--name", builder_name, builder_name_with_arch(local_arch), "--platform", "linux/#{local_arch}"
     end
 
     def append_remote_buildx
-      docker :buildx, :create, "--append", "--name", builder_name, builder_name_with_arch(remote["arch"]), "--platform", "linux/#{remote["arch"]}"
+      docker :buildx, :create, "--append", "--name", builder_name, builder_name_with_arch(remote_arch), "--platform", "linux/#{remote_arch}"
     end
 
     def create_contexts
       combine \
-        create_context(local["arch"], local["host"]),
+        create_context(local_arch, local_host),
-        create_context(remote["arch"], remote["host"])
+        create_context(remote_arch, remote_host)
     end
 
     def create_context(arch, host)
@@ -41,19 +41,11 @@ class Mrsk::Commands::Builder::Multiarch::Remote < Mrsk::Commands::Builder::Mult
 
     def remove_contexts
       combine \
-        remove_context(local["arch"]),
+        remove_context(local_arch),
-        remove_context(remote["arch"])
+        remove_context(remote_arch)
     end
 
     def remove_context(arch)
       docker :context, :rm, builder_name_with_arch(arch)
     end
-
-    def local
-      config.builder["local"]
-    end
-
-    def remote
-      config.builder["remote"]
-    end
 end

lib/mrsk/commands/builder/native.rb

@@ -1,10 +1,10 @@
 class Mrsk::Commands::Builder::Native < Mrsk::Commands::Builder::Base
   def create
-    # No-op on native
+    # No-op on native without cache
   end
 
   def remove
-    # No-op on native
+    # No-op on native without cache
   end
 
   def push

lib/mrsk/commands/builder/native/cached.rb

@@ -0,0 +1,16 @@
+class Mrsk::Commands::Builder::Native::Cached < Mrsk::Commands::Builder::Native
+  def create
+    docker :buildx, :create, "--use", "--driver=docker-container"
+  end
+
+  def remove
+    docker :buildx, :rm, builder_name
+  end
+
+  def push
+    docker :buildx, :build,
+      "--push",
+      *build_options,
+      build_context
+  end
+end

lib/mrsk/commands/builder/native/remote.rb

@@ -28,29 +28,21 @@ class Mrsk::Commands::Builder::Native::Remote < Mrsk::Commands::Builder::Native
 
 
   private
-    def arch
-      config.builder["remote"]["arch"]
-    end
-
-    def host
-      config.builder["remote"]["host"]
-    end
-
     def builder_name
       "mrsk-#{config.service}-native-remote"
     end
 
     def builder_name_with_arch
-      "#{builder_name}-#{arch}"
+      "#{builder_name}-#{remote_arch}"
     end
 
     def platform
-      "linux/#{arch}"
+      "linux/#{remote_arch}"
     end
 
     def create_context
       docker :context, :create,
-        builder_name_with_arch, "--description", "'#{builder_name} #{arch} native host'", "--docker", "'host=#{host}'"
+        builder_name_with_arch, "--description", "'#{builder_name} #{remote_arch} native host'", "--docker", "'host=#{remote_host}'"
     end
 
     def remove_context

lib/mrsk/commands/docker.rb

@@ -0,0 +1,21 @@
+class Mrsk::Commands::Docker < Mrsk::Commands::Base
+  # Install Docker using the https://github.com/docker/docker-install convenience script.
+  def install
+    pipe [ :curl, "-fsSL", "https://get.docker.com" ], :sh
+  end
+
+  # Checks the Docker client version. Fails if Docker is not installed.
+  def installed?
+    docker "-v"
+  end
+
+  # Checks the Docker server version. Fails if Docker is not running.
+  def running?
+    docker :version
+  end
+
+  # Do we have superuser access to install Docker and start system services?
+  def superuser?
+    [ '[ "${EUID:-$(id -u)}" -eq 0 ]' ]
+  end
+end

lib/mrsk/commands/healthcheck.rb

@@ -11,14 +11,19 @@ class Mrsk::Commands::Healthcheck < Mrsk::Commands::Base
       "--label", "service=#{container_name}",
       "-e", "MRSK_CONTAINER_NAME=\"#{container_name}\"",
       *web.env_args,
+      *web.health_check_args,
       *config.volume_args,
       *web.option_args,
       config.absolute_image,
       web.cmd
   end
 
-  def curl
+  def status
-    [ :curl, "--silent", "--output", "/dev/null", "--write-out", "'%{http_code}'", "--max-time", "2", health_url ]
+    pipe container_id, xargs(docker(:inspect, "--format", DOCKER_HEALTH_STATUS_FORMAT))
+  end
+
+  def container_health_log
+    pipe container_id, xargs(docker(:inspect, "--format", DOCKER_HEALTH_LOG_FORMAT))
   end
 
   def logs

lib/mrsk/commands/hook.rb

@@ -0,0 +1,14 @@
+class Mrsk::Commands::Hook < Mrsk::Commands::Base
+  def run(hook, **details)
+    [ hook_file(hook), env: tags(**details).env ]
+  end
+
+  def hook_exists?(hook)
+    Pathname.new(hook_file(hook)).exist?
+  end
+
+  private
+    def hook_file(hook)
+      "#{config.hooks_path}/#{hook}"
+    end
+end

lib/mrsk/commands/lock.rb

@@ -1,5 +1,5 @@
 require "active_support/duration"
-require "active_support/core_ext/numeric/time"
+require "time"
 
 class Mrsk::Commands::Lock < Mrsk::Commands::Base
   def acquire(message, version)
@@ -40,7 +40,7 @@ class Mrsk::Commands::Lock < Mrsk::Commands::Base
     end
 
     def lock_dir
-      :mrsk_lock
+      "mrsk_lock-#{config.service}"
     end
 
     def lock_details_file
@@ -49,7 +49,7 @@ class Mrsk::Commands::Lock < Mrsk::Commands::Base
 
     def lock_details(message, version)
       <<~DETAILS.strip
-        Locked by: #{locked_by} at #{Time.now.gmtime}
+        Locked by: #{locked_by} at #{Time.now.utc.iso8601}
         Version: #{version}
         Message: #{message}
       DETAILS

lib/mrsk/commands/prune.rb

@@ -2,11 +2,37 @@ require "active_support/duration"
 require "active_support/core_ext/numeric/time"
 
 class Mrsk::Commands::Prune < Mrsk::Commands::Base
-  def images(until_hours: 7.days.in_hours.to_i)
+  def dangling_images
-    docker :image, :prune, "--all", "--force", "--filter", "label=service=#{config.service}", "--filter", "until=#{until_hours}h"
+    docker :image, :prune, "--force", "--filter", "label=service=#{config.service}", "--filter", "dangling=true"
   end
 
-  def containers(until_hours: 3.days.in_hours.to_i)
+  def tagged_images
-    docker :container, :prune, "--force", "--filter", "label=service=#{config.service}", "--filter", "until=#{until_hours}h"
+    pipe \
+      docker(:image, :ls, *service_filter, "--format", "'{{.ID}} {{.Repository}}:{{.Tag}}'"),
+      "grep -v -w \"#{active_image_list}\"",
+      "while read image tag; do docker rmi $tag; done"
   end
+
+  def containers(keep_last: 5)
+    pipe \
+      docker(:ps, "-q", "-a", *service_filter, *stopped_containers_filters),
+      "tail -n +#{keep_last + 1}",
+      "while read container_id; do docker rm $container_id; done"
+  end
+
+  private
+    def stopped_containers_filters
+      [ "created", "exited", "dead" ].flat_map { |status| ["--filter", "status=#{status}"] }
+    end
+
+    def active_image_list
+      # Pull the images that are used by any containers
+      # Append repo:latest - to avoid deleting the latest tag
+      # Append repo:<none> - to avoid deleting dangling images that are in use. Unused dangling images are deleted separately
+      "$(docker container ls -a --format '{{.Image}}\\|' --filter label=service=#{config.service} | tr -d '\\n')#{config.latest_image}\\|#{config.repository}:<none>"
+    end
+
+    def service_filter
+      [ "--filter", "label=service=#{config.service}" ]
+    end
 end

lib/mrsk/commands/traefik.rb

@@ -1,8 +1,11 @@
 class Mrsk::Commands::Traefik < Mrsk::Commands::Base
-  delegate :argumentize, :optionize, to: Mrsk::Utils
+  delegate :argumentize, :argumentize_env_with_secrets, :optionize, to: Mrsk::Utils
 
   DEFAULT_IMAGE = "traefik:v2.9"
   CONTAINER_PORT = 80
+  DEFAULT_ARGS = {
+    'log.level' => 'DEBUG'
+  }
 
   def run
     docker :run, "--name traefik",
@@ -10,12 +13,12 @@ class Mrsk::Commands::Traefik < Mrsk::Commands::Base
       "--restart", "unless-stopped",
       "--publish", port,
       "--volume", "/var/run/docker.sock:/var/run/docker.sock",
+      *env_args,
       *config.logging_args,
       *label_args,
       *docker_options_args,
       image,
       "--providers.docker",
-      "--log.level=DEBUG",
       *cmd_option_args
   end
 
@@ -61,6 +64,16 @@ class Mrsk::Commands::Traefik < Mrsk::Commands::Base
       argumentize "--label", labels
     end
 
+    def env_args
+      env_config = config.traefik["env"] || {}
+
+      if env_config.present?
+        argumentize_env_with_secrets(env_config)
+      else
+        []
+      end
+    end
+
     def labels
       config.traefik["labels"] || []
     end
@@ -75,9 +88,9 @@ class Mrsk::Commands::Traefik < Mrsk::Commands::Base
 
     def cmd_option_args
       if args = config.traefik["args"]
-        optionize args, with: "="
+        optionize DEFAULT_ARGS.merge(args), with: "="
       else
-        []
+        optionize DEFAULT_ARGS, with: "="
       end
     end
 

lib/mrsk/configuration.rb

@@ -6,7 +6,7 @@ require "erb"
 require "net/ssh/proxy/jump"
 
 class Mrsk::Configuration
-  delegate :service, :image, :servers, :env, :labels, :registry, :builder, :stop_wait_time, to: :raw_config, allow_nil: true
+  delegate :service, :image, :servers, :env, :labels, :registry, :stop_wait_time, :hooks_path, to: :raw_config, allow_nil: true
   delegate :argumentize, :argumentize_env_with_secrets, :optionize, to: Mrsk::Utils
 
   attr_accessor :destination
@@ -50,7 +50,7 @@ class Mrsk::Configuration
   end
 
   def version
-    @declared_version.presence || ENV["VERSION"] || current_commit_hash
+    @declared_version.presence || ENV["VERSION"] || git_version
   end
 
   def abbreviated_version
@@ -87,6 +87,10 @@ class Mrsk::Configuration
     roles.select(&:running_traefik?).flat_map(&:hosts).uniq
   end
 
+  def boot
+    Mrsk::Configuration::Boot.new(config: self)
+  end
+
 
   def repository
     [ raw_config.registry["server"], image ].compact.join("/")
@@ -149,14 +153,18 @@ class Mrsk::Configuration
   end
 
   def ssh_options
-    { user: ssh_user, proxy: ssh_proxy, auth_methods: [ "publickey" ] }.compact
+    { user: ssh_user, proxy: ssh_proxy, auth_methods: [ "publickey" ], logger: ssh_logger }.compact
   end
 
-
+  def ssh_logger
-  def audit_broadcast_cmd
+    @ssh_logger ||= ::Logger.new(STDERR).tap { |logger| logger.level = ssh_log_level }
-    raw_config.audit_broadcast_cmd
   end
 
+  def ssh_log_level
+    (raw_config.ssh && raw_config.ssh["log_level"]) || ::Logger::FATAL
+  end
+
+
   def healthcheck
     { "path" => "/up", "port" => 3000, "max_attempts" => 7 }.merge(raw_config.healthcheck || {})
   end
@@ -165,8 +173,12 @@ class Mrsk::Configuration
     raw_config.readiness_delay || 7
   end
 
+  def minimum_version
+    raw_config.minimum_version
+  end
+
   def valid?
-    ensure_required_keys_present && ensure_env_available
+    ensure_required_keys_present && ensure_valid_mrsk_version
   end
 
 
@@ -181,8 +193,9 @@ class Mrsk::Configuration
       service_with_version: service_with_version,
       env_args: env_args,
       volume_args: volume_args,
-      ssh_options: ssh_options,
+      ssh_options: ssh_options.except(:logger),
-      builder: raw_config.builder,
+      ssh_log_level: ssh_log_level,
+      builder: builder.to_h,
       accessories: raw_config.accessories,
       logging: logging_args,
       healthcheck: healthcheck
@@ -193,6 +206,22 @@ class Mrsk::Configuration
     raw_config.traefik || {}
   end
 
+  def hooks_path
+    raw_config.hooks_path || ".mrsk/hooks"
+  end
+
+  def builder
+    Mrsk::Configuration::Builder.new(config: self)
+  end
+
+  # Will raise KeyError if any secret ENVs are missing
+  def ensure_env_available
+    env_args
+    roles.each(&:env_args)
+
+    true
+  end
+
   private
     # Will raise ArgumentError if any required config keys are missing
     def ensure_required_keys_present
@@ -217,22 +246,25 @@ class Mrsk::Configuration
       true
     end
 
-    # Will raise KeyError if any secret ENVs are missing
+    def ensure_valid_mrsk_version
-    def ensure_env_available
+      if minimum_version && Gem::Version.new(minimum_version) > Gem::Version.new(Mrsk::VERSION)
-      env_args
+        raise ArgumentError, "Current version is #{Mrsk::VERSION}, minimum required is #{minimum_version}"
-      roles.each(&:env_args)
+      end
 
       true
     end
 
+
     def role_names
       raw_config.servers.is_a?(Array) ? [ "web" ] : raw_config.servers.keys.sort
     end
 
-    def current_commit_hash
+    def git_version
-      @current_commit_hash ||=
+      @git_version ||=
         if system("git rev-parse")
-          `git rev-parse HEAD`.strip
+          uncommitted_suffix = `git status --porcelain`.strip.present? ? "_uncommitted_#{SecureRandom.hex(8)}" : ""
+
+          "#{`git rev-parse HEAD`.strip}#{uncommitted_suffix}"
         else
           raise "Can't use commit hash as version, no git repository found in #{Dir.pwd}"
         end

lib/mrsk/configuration/boot.rb

@@ -0,0 +1,20 @@
+class Mrsk::Configuration::Boot
+  def initialize(config:)
+    @options = config.raw_config.boot || {}
+    @host_count = config.all_hosts.count
+  end
+
+  def limit
+    limit = @options["limit"]
+
+    if limit.to_s.end_with?("%")
+      @host_count * limit.to_i / 100
+    else
+      limit
+    end
+  end
+
+  def wait
+    @options["wait"]
+  end
+end

lib/mrsk/configuration/builder.rb

@@ -0,0 +1,114 @@
+class Mrsk::Configuration::Builder
+  def initialize(config:)
+    @options = config.raw_config.builder || {}
+    @image = config.image
+    @server = config.registry["server"]
+
+    valid?
+  end
+
+  def to_h
+    @options
+  end
+
+  def multiarch?
+    @options["multiarch"] != false
+  end
+
+  def local?
+    !!@options["local"]
+  end
+
+  def remote?
+    !!@options["remote"]
+  end
+
+  def cached?
+    !!@options["cache"]
+  end
+
+  def args
+    @options["args"] || {}
+  end
+
+  def secrets
+    @options["secrets"] || []
+  end
+
+  def dockerfile
+    @options["dockerfile"] || "Dockerfile"
+  end
+
+  def context
+    @options["context"] || "."
+  end
+
+  def local_arch
+    @options["local"]["arch"] if local?
+  end
+
+  def local_host
+    @options["local"]["host"] if local?
+  end
+
+  def remote_arch
+    @options["remote"]["arch"] if remote?
+  end
+
+  def remote_host
+    @options["remote"]["host"] if remote?
+  end
+
+  def cache_from
+    if cached?
+      case @options["cache"]["type"]
+      when "gha"
+        cache_from_config_for_gha
+      when "registry"
+        cache_from_config_for_registry
+      end
+    end
+  end
+
+  def cache_to
+    if cached?
+      case @options["cache"]["type"]
+      when "gha"
+        cache_to_config_for_gha
+      when "registry"
+        cache_to_config_for_registry
+      end
+    end
+  end
+
+  private
+    def valid?
+      if @options["local"] && !@options["remote"]
+        raise ArgumentError, "You must specify both local and remote builder config for remote multiarch builds"
+      end
+
+      if @options["cache"] && @options["cache"]["type"]
+        raise ArgumentError, "Invalid cache type: #{@options["cache"]["type"]}" unless ["gha", "registry"].include?(@options["cache"]["type"])
+      end
+    end
+
+    def cache_image
+      @options["cache"]&.fetch("image", nil) || "#{@image}-build-cache"
+    end
+
+    def cache_from_config_for_gha
+      "type=gha"
+    end
+
+    def cache_from_config_for_registry
+      [ "type=registry", "ref=#{@server}/#{cache_image}" ].compact.join(",")
+    end
+
+    def cache_to_config_for_gha
+      [ "type=gha", @options["cache"]&.fetch("options", nil)].compact.join(",")
+    end
+
+    def cache_to_config_for_registry
+      [ "type=registry", @options["cache"]&.fetch("options", nil), "ref=#{@server}/#{cache_image}" ].compact.join(",")
+    end
+end

lib/mrsk/configuration/role.rb

@@ -35,6 +35,28 @@ class Mrsk::Configuration::Role
     argumentize_env_with_secrets env
   end
 
+  def health_check_args
+    if health_check_cmd.present?
+      optionize({ "health-cmd" => health_check_cmd, "health-interval" => health_check_interval })
+    else
+      []
+    end
+  end
+
+  def health_check_cmd
+    options = specializations["healthcheck"] || {}
+    options = config.healthcheck.merge(options) if running_traefik?
+
+    options["cmd"] || http_health_check(port: options["port"], path: options["path"])
+  end
+
+  def health_check_interval
+    options = specializations["healthcheck"] || {}
+    options = config.healthcheck.merge(options) if running_traefik?
+
+    options["interval"] || "1s"
+  end
+
   def cmd
     specializations["cmd"]
   end
@@ -74,9 +96,10 @@ class Mrsk::Configuration::Role
     def traefik_labels
       if running_traefik?
         {
+          # Setting a service property ensures that the generated service name will be consistent between versions
+          "traefik.http.services.#{traefik_service}.loadbalancer.server.scheme" => "http",
+
           "traefik.http.routers.#{traefik_service}.rule" => "PathPrefix(`/`)",
-          "traefik.http.services.#{traefik_service}.loadbalancer.healthcheck.path" => config.healthcheck["path"],
-          "traefik.http.services.#{traefik_service}.loadbalancer.healthcheck.interval" => "1s",
           "traefik.http.middlewares.#{traefik_service}-retry.retry.attempts" => "5",
           "traefik.http.middlewares.#{traefik_service}-retry.retry.initialinterval" => "500ms",
           "traefik.http.routers.#{traefik_service}.middlewares" => "#{traefik_service}-retry@docker"
@@ -125,4 +148,8 @@ class Mrsk::Configuration::Role
         new_env["clear"] = (clear_app_env + clear_role_env).uniq
       end
     end
+
+    def http_health_check(port:, path:)
+      "curl -f #{URI.join("http://localhost:#{port}", path)} || exit 1" if path.present? || port.present?
+    end
 end

lib/mrsk/sshkit_with_ext.rb

@@ -1,12 +1,56 @@
 require "sshkit"
 require "sshkit/dsl"
+require "active_support/core_ext/hash/deep_merge"
+require "json"
 
 class SSHKit::Backend::Abstract
-  def capture_with_info(*args)
+  def capture_with_info(*args, **kwargs)
-    capture(*args, verbosity: Logger::INFO)
+    capture(*args, **kwargs, verbosity: Logger::INFO)
+  end
+
+  def capture_with_debug(*args, **kwargs)
+    capture(*args, **kwargs, verbosity: Logger::DEBUG)
+  end
+
+  def capture_with_pretty_json(*args, **kwargs)
+    JSON.pretty_generate(JSON.parse(capture(*args, **kwargs)))
   end
 
   def puts_by_host(host, output, type: "App")
     puts "#{type} Host: #{host}\n#{output}\n\n"
   end
+
+  # Our execution pattern is for the CLI execute args lists returned
+  # from commands, but this doesn't support returning execution options
+  # from the command.
+  #
+  # Support this by using kwargs for CLI options and merging with the
+  # args-extracted options.
+  module CommandEnvMerge
+    private
+
+    # Override to merge options returned by commands in the args list with
+    # options passed by the CLI and pass them along as kwargs.
+    def command(args, options)
+      more_options, args = args.partition { |a| a.is_a? Hash }
+      more_options << options
+
+      build_command(args, **more_options.reduce(:deep_merge))
+    end
+
+    # Destructure options to pluck out env for merge
+    def build_command(args, env: nil, **options)
+      # Rely on native Ruby kwargs precedence rather than explicit Hash merges
+      SSHKit::Command.new(*args, **default_command_options, **options, env: env_for(env))
+    end
+
+    def default_command_options
+      { in: pwd_path, host: @host, user: @user, group: @group }
+    end
+
+    def env_for(env)
+      @env.to_h.merge(env.to_h)
+    end
+  end
+  prepend CommandEnvMerge
 end

lib/mrsk/tags.rb

@@ -0,0 +1,39 @@
+require "time"
+
+class Mrsk::Tags
+  attr_reader :config, :tags
+
+  class << self
+    def from_config(config, **extra)
+      new(**default_tags(config), **extra)
+    end
+
+    def default_tags(config)
+      { recorded_at: Time.now.utc.iso8601,
+        performer: `whoami`.chomp,
+        destination: config.destination,
+        version: config.version,
+        service_version: service_version(config) }
+    end
+
+    def service_version(config)
+      [ config.service, config.abbreviated_version ].compact.join("@")
+    end
+  end
+
+  def initialize(**tags)
+    @tags = tags.compact
+  end
+
+  def env
+    tags.transform_keys { |detail| "MRSK_#{detail.upcase}" }
+  end
+
+  def to_s
+    tags.values.map { |value| "[#{value}]" }.join(" ")
+  end
+
+  def except(*tags)
+    self.class.new(**self.tags.except(*tags))
+  end
+end

lib/mrsk/utils.rb

@@ -1,6 +1,8 @@
 module Mrsk::Utils
   extend self
 
+  DOLLAR_SIGN_WITHOUT_SHELL_EXPANSION_REGEX = /\$(?!{[^\}]*\})/
+
   # Return a list of escaped shell arguments using the same named argument against the passed attributes (hash or array).
   def argumentize(argument, attributes, sensitive: false)
     Array(attributes).flat_map do |key, value|
@@ -75,11 +77,20 @@ module Mrsk::Utils
 
   # Escape a value to make it safe for shell use.
   def escape_shell_value(value)
-    value.to_s.dump.gsub(/`/, '\\\\`')
+    value.to_s.dump
+      .gsub(/`/, '\\\\`')
+      .gsub(DOLLAR_SIGN_WITHOUT_SHELL_EXPANSION_REGEX, '\$')
   end
 
   # Abbreviate a git revhash for concise display
   def abbreviate_version(version)
-    version[0...7] if version
+    if version
+      # Don't abbreviate <sha>_uncommitted_<etc>
+      if version.include?("_")
+        version
+      else
+        version[0...7]
+      end
+    end
   end
 end

lib/mrsk/utils/healthcheck_poller.rb

@@ -0,0 +1,39 @@
+class Mrsk::Utils::HealthcheckPoller
+  TRAEFIK_HEALTHY_DELAY = 2
+
+  class HealthcheckError < StandardError; end
+
+  class << self
+    def wait_for_healthy(pause_after_ready: false, &block)
+      attempt = 1
+      max_attempts = MRSK.config.healthcheck["max_attempts"]
+
+      begin
+        case status = block.call
+        when "healthy"
+          sleep TRAEFIK_HEALTHY_DELAY if pause_after_ready
+        when "running" # No health check configured
+          sleep MRSK.config.readiness_delay if pause_after_ready
+        else
+          raise HealthcheckError, "container not ready (#{status})"
+        end
+      rescue HealthcheckError => e
+        if attempt <= max_attempts
+          info "#{e.message}, retrying in #{attempt}s (attempt #{attempt}/#{max_attempts})..."
+          sleep attempt
+          attempt += 1
+          retry
+        else
+          raise
+        end
+      end
+
+      info "Container is healthy!"
+    end
+
+    private
+      def info(message)
+        SSHKit.config.output.info(message)
+      end
+  end
+end

lib/mrsk/version.rb

@@ -1,3 +1,3 @@
 module Mrsk
-  VERSION = "0.11.0"
+  VERSION = "0.15.0"
 end

test/cli/accessory_test.rb

@@ -40,9 +40,10 @@ class CliAccessoryTest < CliTestCase
   end
 
   test "reboot" do
+    Mrsk::Commands::Registry.any_instance.expects(:login)
     Mrsk::Cli::Accessory.any_instance.expects(:stop).with("mysql")
     Mrsk::Cli::Accessory.any_instance.expects(:remove_container).with("mysql")
-    Mrsk::Cli::Accessory.any_instance.expects(:boot).with("mysql")
+    Mrsk::Cli::Accessory.any_instance.expects(:boot).with("mysql", login: false)
 
     run_command("reboot", "mysql")
   end

test/cli/app_test.rb

@@ -2,11 +2,10 @@ require_relative "cli_test_case"
 
 class CliAppTest < CliTestCase
   test "boot" do
-    # Stub current version fetch
+    stub_running
-    SSHKit::Backend::Abstract.any_instance.stubs(:capture).returns("123") # old version
-
     run_command("boot").tap do |output|
-      assert_match "docker run --detach --restart unless-stopped", output
+      assert_match "docker tag dhh/app:latest dhh/app:latest", output
+      assert_match /docker run --detach --restart unless-stopped --name app-web-latest --hostname 1.1.1.1-[0-9a-f]{12} /, output
       assert_match "docker container ls --all --filter name=^app-web-123$ --quiet | xargs docker stop", output
     end
   end
@@ -15,23 +14,49 @@ class CliAppTest < CliTestCase
     run_command("details") # Preheat MRSK const
 
     SSHKit::Backend::Abstract.any_instance.expects(:capture_with_info)
-      .with(:docker, :container, :ls, "--all", "--filter", "name=^app-web-latest$", "--quiet")
+      .with(:docker, :container, :ls, "--filter", "name=^app-web-latest$", "--quiet", raise_on_non_zero_exit: false)
       .returns("12345678") # running version
 
     SSHKit::Backend::Abstract.any_instance.expects(:capture_with_info)
-      .with(:docker, :ps, "--filter", "label=service=app", "--filter", "label=role=web", "--format", "\"{{.Names}}\"", "|", "sed 's/-/\\n/g'", "|", "tail -n 1")
+      .with(:docker, :container, :ls, "--all", "--filter", "name=^app-web-latest$", "--quiet", "|", :xargs, :docker, :inspect, "--format", "'{{if .State.Health}}{{.State.Health.Status}}{{else}}{{.State.Status}}{{end}}'")
+      .returns("running") # health check
+
+    SSHKit::Backend::Abstract.any_instance.expects(:capture_with_info)
+      .with(:docker, :ps, "--filter", "label=service=app", "--filter", "label=role=web", "--filter", "status=running", "--filter", "status=restarting", "--latest", "--format", "\"{{.Names}}\"", "|", "grep -oE \"\\-[^-]+$\"", "|", "cut -c 2-", raise_on_non_zero_exit: false)
       .returns("123") # old version
 
     run_command("boot").tap do |output|
       assert_match /Renaming container .* to .* as already deployed on 1.1.1.1/, output # Rename
-      assert_match /docker rename .* .*/, output
+      assert_match /docker rename app-web-latest app-web-latest_replaced_[0-9a-f]{16}/, output
-      assert_match "docker run --detach --restart unless-stopped", output
+      assert_match /docker run --detach --restart unless-stopped --name app-web-latest --hostname 1.1.1.1-[0-9a-f]{12} /, output
       assert_match "docker container ls --all --filter name=^app-web-123$ --quiet | xargs docker stop", output
     end
   ensure
     Thread.report_on_exception = true
   end
 
+  test "boot uses group strategy when specified" do
+    Mrsk::Cli::App.any_instance.stubs(:on).with("1.1.1.1").twice # acquire & release lock
+    Mrsk::Cli::App.any_instance.stubs(:on).with([ "1.1.1.1" ]) # tag container
+
+    # Strategy is used when booting the containers
+    Mrsk::Cli::App.any_instance.expects(:on).with([ "1.1.1.1" ], in: :groups, limit: 3, wait: 2).with_block_given
+
+    run_command("boot", config: :with_boot_strategy)
+  end
+
+  test "boot errors leave lock in place" do
+    invoke_options = { "config_file" => "test/fixtures/deploy_simple.yml", "version" => "999" }
+
+    Mrsk::Cli::App.any_instance.expects(:using_version).raises(RuntimeError)
+
+    assert !MRSK.holding_lock?
+    assert_raises(RuntimeError) do
+      stderred { run_command("boot") }
+    end
+    assert MRSK.holding_lock?
+  end
+
   test "start" do
     run_command("start").tap do |output|
       assert_match "docker start app-web-999", output
@@ -40,7 +65,28 @@ class CliAppTest < CliTestCase
 
   test "stop" do
     run_command("stop").tap do |output|
-      assert_match "docker ps --quiet --filter label=service=app --filter label=role=web | xargs docker stop", output
+      assert_match "docker ps --quiet --filter label=service=app --filter label=role=web --filter status=running --filter status=restarting --latest | xargs docker stop", output
+    end
+  end
+
+  test "stale_containers" do
+    SSHKit::Backend::Abstract.any_instance.expects(:capture_with_info)
+      .with(:docker, :ps, "--filter", "label=service=app", "--filter", "label=role=web", "--format", "\"{{.Names}}\"", "|", "grep -oE \"\\-[^-]+$\"", "|", "cut -c 2-", raise_on_non_zero_exit: false)
+      .returns("12345678\n87654321")
+
+    run_command("stale_containers").tap do |output|
+      assert_match /Detected stale container for role web with version 87654321/, output
+    end
+  end
+
+  test "stop stale_containers" do
+    SSHKit::Backend::Abstract.any_instance.expects(:capture_with_info)
+      .with(:docker, :ps, "--filter", "label=service=app", "--filter", "label=role=web", "--format", "\"{{.Names}}\"", "|", "grep -oE \"\\-[^-]+$\"", "|", "cut -c 2-", raise_on_non_zero_exit: false)
+      .returns("12345678\n87654321")
+
+    run_command("stale_containers", "--stop").tap do |output|
+      assert_match /Stopping stale container for role web with version 87654321/, output
+      assert_match /#{Regexp.escape("docker container ls --all --filter name=^app-web-87654321$ --quiet | xargs docker stop")}/, output
     end
   end
 
@@ -52,7 +98,7 @@ class CliAppTest < CliTestCase
 
   test "remove" do
     run_command("remove").tap do |output|
-      assert_match /#{Regexp.escape("docker ps --quiet --filter label=service=app --filter label=role=web | xargs docker stop")}/, output
+      assert_match /#{Regexp.escape("docker ps --quiet --filter label=service=app --filter label=role=web --filter status=running --filter status=restarting --latest | xargs docker stop")}/, output
       assert_match /#{Regexp.escape("docker container prune --force --filter label=service=app")}/, output
       assert_match /#{Regexp.escape("docker image prune --all --force --filter label=service=app")}/, output
     end
@@ -84,7 +130,7 @@ class CliAppTest < CliTestCase
 
   test "exec with reuse" do
     run_command("exec", "--reuse", "ruby -v").tap do |output|
-      assert_match "docker ps --filter label=service=app --format \"{{.Names}}\" | sed 's/-/\\n/g' | tail -n 1", output # Get current version
+      assert_match "docker ps --filter label=service=app --filter status=running --filter status=restarting --latest --format \"{{.Names}}\" | grep -oE \"\\-[^-]+$\" | cut -c 2-", output # Get current version
       assert_match "docker exec app-web-999 ruby -v", output
     end
   end
@@ -103,33 +149,41 @@ class CliAppTest < CliTestCase
 
   test "logs" do
     SSHKit::Backend::Abstract.any_instance.stubs(:exec)
-      .with("ssh -t root@1.1.1.1 'docker ps --quiet --filter label=service=app --filter label=role=web | xargs docker logs --timestamps --tail 10 2>&1'")
+      .with("ssh -t root@1.1.1.1 'docker ps --quiet --filter label=service=app --filter label=role=web --filter status=running --filter status=restarting --latest| xargs docker logs --timestamps --tail 10 2>&1'")
 
-    assert_match "docker ps --quiet --filter label=service=app --filter label=role=web | xargs docker logs --tail 100 2>&1", run_command("logs")
+    assert_match "docker ps --quiet --filter label=service=app --filter label=role=web --filter status=running --filter status=restarting --latest | xargs docker logs --tail 100 2>&1", run_command("logs")
   end
 
   test "logs with follow" do
     SSHKit::Backend::Abstract.any_instance.stubs(:exec)
-      .with("ssh -t root@1.1.1.1 'docker ps --quiet --filter label=service=app --filter label=role=web | xargs docker logs --timestamps --tail 10 --follow 2>&1'")
+      .with("ssh -t root@1.1.1.1 'docker ps --quiet --filter label=service=app --filter label=role=web --filter status=running --filter status=restarting --latest | xargs docker logs --timestamps --tail 10 --follow 2>&1'")
 
-    assert_match "docker ps --quiet --filter label=service=app --filter label=role=web | xargs docker logs --timestamps --tail 10 --follow 2>&1", run_command("logs", "--follow")
+    assert_match "docker ps --quiet --filter label=service=app --filter label=role=web --filter status=running --filter status=restarting --latest | xargs docker logs --timestamps --tail 10 --follow 2>&1", run_command("logs", "--follow")
   end
 
   test "version" do
     run_command("version").tap do |output|
-      assert_match "docker ps --filter label=service=app --format \"{{.Names}}\" | sed 's/-/\\n/g' | tail -n 1", output
+      assert_match "docker ps --filter label=service=app --filter status=running --filter status=restarting --latest --format \"{{.Names}}\" | grep -oE \"\\-[^-]+$\" | cut -c 2-", output
     end
   end
 
 
   test "version through main" do
     stdouted { Mrsk::Cli::Main.start(["app", "version", "-c", "test/fixtures/deploy_with_accessories.yml", "--hosts", "1.1.1.1"]) }.tap do |output|
-      assert_match "docker ps --filter label=service=app --format \"{{.Names}}\" | sed 's/-/\\n/g' | tail -n 1", output
+      assert_match "docker ps --filter label=service=app --filter status=running --filter status=restarting --latest --format \"{{.Names}}\" | grep -oE \"\\-[^-]+$\" | cut -c 2-", output
     end
   end
 
   private
-    def run_command(*command)
+    def run_command(*command, config: :with_accessories)
-      stdouted { Mrsk::Cli::App.start([*command, "-c", "test/fixtures/deploy_with_accessories.yml", "--hosts", "1.1.1.1"]) }
+      stdouted { Mrsk::Cli::App.start([*command, "-c", "test/fixtures/deploy_#{config}.yml", "--hosts", "1.1.1.1"]) }
+    end
+
+    def stub_running
+      SSHKit::Backend::Abstract.any_instance.stubs(:capture_with_info).returns("123") # old version
+
+      SSHKit::Backend::Abstract.any_instance.expects(:capture_with_info)
+        .with(:docker, :container, :ls, "--all", "--filter", "name=^app-web-latest$", "--quiet", "|", :xargs, :docker, :inspect, "--format", "'{{if .State.Health}}{{.State.Health.Status}}{{else}}{{.State.Status}}{{end}}'")
+        .returns("running") # health check
     end
 end

test/cli/build_test.rb

@@ -9,7 +9,12 @@ class CliBuildTest < CliTestCase
   end
 
   test "push" do
+    Mrsk::Commands::Hook.any_instance.stubs(:hook_exists?).returns(true)
+    hook_variables = { version: 999, service_version: "app@999", hosts: "1.1.1.1,1.1.1.2,1.1.1.3,1.1.1.4", command: "build", subcommand: "push" }
+
     run_command("push").tap do |output|
+      assert_hook_ran "pre-build", output, **hook_variables
+      assert_match /docker --version && docker buildx version/, output
       assert_match /docker buildx build --push --platform linux\/amd64,linux\/arm64 --builder mrsk-app-multiarch -t dhh\/app:999 -t dhh\/app:latest --label service="app" --file Dockerfile \. as .*@localhost/, output
     end
   end
@@ -17,7 +22,10 @@ class CliBuildTest < CliTestCase
   test "push without builder" do
     stub_locking
     SSHKit::Backend::Abstract.any_instance.stubs(:execute)
-      .with { |arg| arg == :docker }
+      .with(:docker, "--version", "&&", :docker, :buildx, "version")
+
+    SSHKit::Backend::Abstract.any_instance.stubs(:execute)
+      .with { |*args| args[0..1] == [:docker, :buildx] }
       .raises(SSHKit::Command::Failed.new("no builder"))
       .then
       .returns(true)
@@ -27,10 +35,28 @@ class CliBuildTest < CliTestCase
     end
   end
 
+  test "push with no buildx plugin" do
+    stub_locking
+    SSHKit::Backend::Abstract.any_instance.stubs(:execute)
+      .with(:docker, "--version", "&&", :docker, :buildx, "version")
+      .raises(SSHKit::Command::Failed.new("no buildx"))
+
+    Mrsk::Commands::Builder.any_instance.stubs(:native_and_local?).returns(false)
+    assert_raises(Mrsk::Cli::Build::BuildError) { run_command("push") }
+  end
+
+  test "push pre-build hook failure" do
+    fail_hook("pre-build")
+
+    assert_raises(Mrsk::Cli::HookError) { run_command("push") }
+
+    assert @executions.none? { |args| args[0..2] == [:docker, :buildx, :build] }
+  end
+
   test "pull" do
     run_command("pull").tap do |output|
       assert_match /docker image rm --force dhh\/app:999/, output
-      assert_match /docker pull dhh\/app:latest/, output
+      assert_match /docker pull dhh\/app:999/, output
     end
   end
 
@@ -73,10 +99,10 @@ class CliBuildTest < CliTestCase
       stdouted { Mrsk::Cli::Build.start([*command, "-c", "test/fixtures/deploy_with_accessories.yml"]) }
     end
 
-    def stub_locking
+    def stub_dependency_checks
       SSHKit::Backend::Abstract.any_instance.stubs(:execute)
-        .with { |arg1, arg2| arg1 == :mkdir && arg2 == :mrsk_lock }
+        .with(:docker, "--version", "&&", :docker, :buildx, "version")
       SSHKit::Backend::Abstract.any_instance.stubs(:execute)
-        .with { |arg1, arg2| arg1 == :rm && arg2 == "mrsk_lock/details" }
+        .with { |*args| args[0..1] == [:docker, :buildx] }
     end
 end

test/cli/cli_test_case.rb

@@ -1,9 +1,6 @@
 require "test_helper"
-require "active_support/testing/stream"
 
 class CliTestCase < ActiveSupport::TestCase
-  include ActiveSupport::Testing::Stream
-
   setup do
     ENV["VERSION"]             = "999"
     ENV["RAILS_MASTER_KEY"]    = "123"
@@ -19,11 +16,41 @@ class CliTestCase < ActiveSupport::TestCase
   end
 
   private
-    def stdouted
+    def fail_hook(hook)
-      capture(:stdout) { yield }.strip
+      @executions = []
+      Mrsk::Commands::Hook.any_instance.stubs(:hook_exists?).returns(true)
+
+      SSHKit::Backend::Abstract.any_instance.stubs(:execute)
+        .with { |*args| @executions << args; args != [".mrsk/hooks/#{hook}"] }
+      SSHKit::Backend::Abstract.any_instance.stubs(:execute)
+        .with { |*args| args.first == ".mrsk/hooks/#{hook}" }
+        .raises(SSHKit::Command::Failed.new("failed"))
     end
 
-    def stderred
+    def stub_locking
-      capture(:stderr) { yield }.strip
+      SSHKit::Backend::Abstract.any_instance.stubs(:execute)
+        .with { |arg1, arg2| arg1 == :mkdir && arg2 == "mrsk_lock-app" }
+      SSHKit::Backend::Abstract.any_instance.stubs(:execute)
+        .with { |arg1, arg2| arg1 == :rm && arg2 == "mrsk_lock-app/details" }
     end
-  end
+
+    def assert_hook_ran(hook, output, version:, service_version:, hosts:, command:, subcommand: nil, runtime: nil)
+      performer = `whoami`.strip
+
+      assert_match "Running the #{hook} hook...\n", output
+
+      expected = %r{Running\s/usr/bin/env\s\.mrsk/hooks/#{hook}\sas\s#{performer}@localhost\n\s
+        DEBUG\s\[[0-9a-f]*\]\sCommand:\s\(\sexport\s
+        MRSK_RECORDED_AT=\"\d\d\d\d-\d\d-\d\dT\d\d:\d\d:\d\dZ\"\s
+        MRSK_PERFORMER=\"#{performer}\"\s
+        MRSK_VERSION=\"#{version}\"\s
+        MRSK_SERVICE_VERSION=\"#{service_version}\"\s
+        MRSK_HOSTS=\"#{hosts}\"\s
+        MRSK_COMMAND=\"#{command}\"\s
+        #{"MRSK_SUBCOMMAND=\\\"#{subcommand}\\\"\\s" if subcommand}
+        #{"MRSK_RUNTIME=\\\"#{runtime}\\\"\\s" if runtime}
+        ;\s/usr/bin/env\s\.mrsk/hooks/#{hook} }x
+
+      assert_match expected, output
+    end
+end

test/cli/healthcheck_test.rb

@@ -5,62 +5,63 @@ class CliHealthcheckTest < CliTestCase
     # Prevent expected failures from outputting to terminal
     Thread.report_on_exception = false
 
-    SSHKit::Backend::Abstract.any_instance.stubs(:sleep) # No sleeping when retrying
+    Mrsk::Utils::HealthcheckPoller.stubs(:sleep) # No sleeping when retrying
+
     SSHKit::Backend::Abstract.any_instance.stubs(:execute)
       .with(:docker, :container, :ls, "--all", "--filter", "name=^healthcheck-app-999$", "--quiet", "|", :xargs, :docker, :stop, raise_on_non_zero_exit: false)
     SSHKit::Backend::Abstract.any_instance.stubs(:execute)
-      .with(:docker, :run, "--detach", "--name", "healthcheck-app-999", "--publish", "3999:3000", "--label", "service=healthcheck-app", "-e", "MRSK_CONTAINER_NAME=\"healthcheck-app\"", "dhh/app:999")
+      .with(:docker, :run, "--detach", "--name", "healthcheck-app-999", "--publish", "3999:3000", "--label", "service=healthcheck-app", "-e", "MRSK_CONTAINER_NAME=\"healthcheck-app\"", "--health-cmd", "\"curl -f http://localhost:3000/up || exit 1\"", "--health-interval", "\"1s\"", "dhh/app:999")
     SSHKit::Backend::Abstract.any_instance.stubs(:execute)
       .with(:docker, :container, :ls, "--all", "--filter", "name=^healthcheck-app-999$", "--quiet", "|", :xargs, :docker, :container, :rm, raise_on_non_zero_exit: false)
 
     # Fail twice to test retry logic
     SSHKit::Backend::Abstract.any_instance.stubs(:capture_with_info)
-      .with(:curl, "--silent", "--output", "/dev/null", "--write-out", "'%{http_code}'", "--max-time", "2", "http://localhost:3999/up")
+      .with(:docker, :container, :ls, "--all", "--filter", "name=^healthcheck-app-999$", "--quiet", "|", :xargs, :docker, :inspect, "--format", "'{{if .State.Health}}{{.State.Health.Status}}{{else}}{{.State.Status}}{{end}}'")
-      .raises(SSHKit::Command::Failed)
+      .returns("starting")
       .then
-      .raises(SSHKit::Command::Failed)
+      .returns("unhealthy")
       .then
-      .returns("200")
+      .returns("healthy")
 
     run_command("perform").tap do |output|
-      assert_match "Health check against /up failed to respond, retrying in 1s (attempt 1/7)...", output
+      assert_match "container not ready (starting), retrying in 1s (attempt 1/7)...", output
-      assert_match "Health check against /up failed to respond, retrying in 2s (attempt 2/7)...", output
+      assert_match "container not ready (unhealthy), retrying in 2s (attempt 2/7)...", output
-      assert_match "Health check against /up succeeded with 200 OK!", output
+      assert_match "Container is healthy!", output
     end
   end
 
-  test "perform failing because of curl" do
+  test "perform failing to become healthy" do
     # Prevent expected failures from outputting to terminal
     Thread.report_on_exception = false
 
-    SSHKit::Backend::Abstract.any_instance.stubs(:execute) # No need to execute anything here
+    Mrsk::Utils::HealthcheckPoller.stubs(:sleep) # No sleeping when retrying
+
+    SSHKit::Backend::Abstract.any_instance.stubs(:execute)
+      .with(:docker, :container, :ls, "--all", "--filter", "name=^healthcheck-app-999$", "--quiet", "|", :xargs, :docker, :stop, raise_on_non_zero_exit: false)
+    SSHKit::Backend::Abstract.any_instance.stubs(:execute)
+      .with(:docker, :run, "--detach", "--name", "healthcheck-app-999", "--publish", "3999:3000", "--label", "service=healthcheck-app", "-e", "MRSK_CONTAINER_NAME=\"healthcheck-app\"", "--health-cmd", "\"curl -f http://localhost:3000/up || exit 1\"", "--health-interval", "\"1s\"", "dhh/app:999")
+    SSHKit::Backend::Abstract.any_instance.stubs(:execute)
+      .with(:docker, :container, :ls, "--all", "--filter", "name=^healthcheck-app-999$", "--quiet", "|", :xargs, :docker, :container, :rm, raise_on_non_zero_exit: false)
+
+    # Continually report unhealthy
     SSHKit::Backend::Abstract.any_instance.stubs(:capture_with_info)
-      .with(:curl, "--silent", "--output", "/dev/null", "--write-out", "'%{http_code}'", "--max-time", "2", "http://localhost:3999/up")
+      .with(:docker, :container, :ls, "--all", "--filter", "name=^healthcheck-app-999$", "--quiet", "|", :xargs, :docker, :inspect, "--format", "'{{if .State.Health}}{{.State.Health.Status}}{{else}}{{.State.Status}}{{end}}'")
-      .returns("curl: command not found")
+      .returns("unhealthy")
+
+    # Capture logs when failing
     SSHKit::Backend::Abstract.any_instance.stubs(:capture_with_info)
       .with(:docker, :container, :ls, "--all", "--filter", "name=^healthcheck-app-999$", "--quiet", "|", :xargs, :docker, :logs, "--tail", 50, "2>&1")
+      .returns("some log output")
 
-    exception = assert_raises SSHKit::Runner::ExecuteError do
+    # Capture container health log when failing
-      run_command("perform")
+    SSHKit::Backend::Abstract.any_instance.stubs(:capture_with_pretty_json)
-    end
+      .with(:docker, :container, :ls, "--all", "--filter", "name=^healthcheck-app-999$", "--quiet", "|", :xargs, :docker, :inspect, "--format",  "'{{json .State.Health}}'")
-    assert_match "Health check against /up failed to return 200 OK!", exception.message
+      .returns('{"Status":"unhealthy","Log":[{"ExitCode": 1,"Output": "/bin/sh: 1: curl: not found\n"}]}"')
-  end
-
-  test "perform failing for unknown reason" do
-    # Prevent expected failures from outputting to terminal
-    Thread.report_on_exception = false
-
-    SSHKit::Backend::Abstract.any_instance.stubs(:execute) # No need to execute anything here
-    SSHKit::Backend::Abstract.any_instance.stubs(:capture_with_info)
-      .with(:curl, "--silent", "--output", "/dev/null", "--write-out", "'%{http_code}'", "--max-time", "2", "http://localhost:3999/up")
-      .returns("500")
-    SSHKit::Backend::Abstract.any_instance.stubs(:capture_with_info)
-      .with(:docker, :container, :ls, "--all", "--filter", "name=^healthcheck-app-999$", "--quiet", "|", :xargs, :docker, :logs, "--tail", 50, "2>&1")
 
     exception = assert_raises do
       run_command("perform")
     end
-    assert_match "Health check against /up failed with status 500", exception.message
+    assert_match "container not ready (unhealthy)", exception.message
   end
 
   private

test/cli/main_test.rb

@@ -10,44 +10,50 @@ class CliMainTest < CliTestCase
   end
 
   test "deploy" do
-    invoke_options = { "config_file" => "test/fixtures/deploy_simple.yml", "skip_broadcast" => false, "version" => "999" }
+    invoke_options = { "config_file" => "test/fixtures/deploy_simple.yml", "version" => "999", "skip_hooks" => false }
 
-    Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:server:bootstrap", [], invoke_options)
     Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:registry:login", [], invoke_options)
     Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:build:deliver", [], invoke_options)
     Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:traefik:boot", [], invoke_options)
     Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:healthcheck:perform", [], invoke_options)
+    Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:app:stale_containers", [], invoke_options)
     Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:app:boot", [], invoke_options)
     Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:prune:all", [], invoke_options)
 
+    Mrsk::Commands::Hook.any_instance.stubs(:hook_exists?).returns(true)
+    hook_variables = { version: 999, service_version: "app@999", hosts: "1.1.1.1,1.1.1.2", command: "deploy" }
+
     run_command("deploy").tap do |output|
-      assert_match /Ensure curl and Docker are installed/, output
+      assert_hook_ran "pre-connect", output, **hook_variables
       assert_match /Log into image registry/, output
       assert_match /Build and push app image/, output
+      assert_hook_ran "pre-deploy", output, **hook_variables
       assert_match /Ensure Traefik is running/, output
       assert_match /Ensure app can pass healthcheck/, output
+      assert_match /Detect stale containers/, output
       assert_match /Prune old containers and images/, output
+      assert_hook_ran "post-deploy", output, **hook_variables, runtime: 0
     end
   end
 
   test "deploy with skip_push" do
-    invoke_options = { "config_file" => "test/fixtures/deploy_simple.yml", "skip_broadcast" => false, "version" => "999" }
+    invoke_options = { "config_file" => "test/fixtures/deploy_simple.yml", "version" => "999", "skip_hooks" => false }
 
-    Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:server:bootstrap", [], invoke_options)
     Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:registry:login", [], invoke_options)
     Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:build:pull", [], invoke_options)
     Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:traefik:boot", [], invoke_options)
     Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:healthcheck:perform", [], invoke_options)
+    Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:app:stale_containers", [], invoke_options)
     Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:app:boot", [], invoke_options)
     Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:prune:all", [], invoke_options)
 
     run_command("deploy", "--skip_push").tap do |output|
       assert_match /Acquiring the deploy lock/, output
-      assert_match /Ensure curl and Docker are installed/, output
       assert_match /Log into image registry/, output
       assert_match /Pull app image/, output
       assert_match /Ensure Traefik is running/, output
       assert_match /Ensure app can pass healthcheck/, output
+      assert_match /Detect stale containers/, output
       assert_match /Prune old containers and images/, output
       assert_match /Releasing the deploy lock/, output
     end
@@ -57,10 +63,11 @@ class CliMainTest < CliTestCase
     Thread.report_on_exception = false
 
     SSHKit::Backend::Abstract.any_instance.stubs(:execute)
-      .with { |*arg| arg[0..1] == [:mkdir, :mrsk_lock] }
+      .with { |*arg| arg[0..1] == [:mkdir, 'mrsk_lock-app'] }
-      .raises(RuntimeError, "mkdir: cannot create directory ‘mrsk_lock’: File exists")
+      .raises(RuntimeError, "mkdir: cannot create directory ‘mrsk_lock-app’: File exists")
 
-    Mrsk::Cli::Base.any_instance.expects(:invoke).with("mrsk:cli:lock:status", [])
+    SSHKit::Backend::Abstract.any_instance.expects(:capture_with_debug)
+      .with(:stat, 'mrsk_lock-app', ">", "/dev/null", "&&", :cat, "mrsk_lock-app/details", "|", :base64, "-d")
 
     assert_raises(Mrsk::Cli::LockError) do
       run_command("deploy")
@@ -71,7 +78,7 @@ class CliMainTest < CliTestCase
     Thread.report_on_exception = false
 
     SSHKit::Backend::Abstract.any_instance.stubs(:execute)
-      .with { |*arg| arg[0..1] == [:mkdir, :mrsk_lock] }
+      .with { |*arg| arg[0..1] == [:mkdir, 'mrsk_lock-app'] }
       .raises(SocketError, "getaddrinfo: nodename nor servname provided, or not known")
 
     assert_raises(SSHKit::Runner::ExecuteError) do
@@ -79,38 +86,70 @@ class CliMainTest < CliTestCase
     end
   end
 
-  test "deploy errors leave lock in place" do
+  test "deploy errors during outside section leave remove lock" do
-    invoke_options = { "config_file" => "test/fixtures/deploy_simple.yml", "skip_broadcast" => false, "version" => "999" }
+    invoke_options = { "config_file" => "test/fixtures/deploy_simple.yml", "version" => "999", "skip_hooks" => false }
 
     Mrsk::Cli::Main.any_instance.expects(:invoke)
-      .with("mrsk:cli:server:bootstrap", [], invoke_options)
+      .with("mrsk:cli:registry:login", [], invoke_options)
       .raises(RuntimeError)
 
-    assert_equal 0, MRSK.lock_count
+    assert !MRSK.holding_lock?
     assert_raises(RuntimeError) do
       stderred { run_command("deploy") }
     end
-    assert_equal 1, MRSK.lock_count
+    assert !MRSK.holding_lock?
+  end
+
+  test "deploy with skipped hooks" do
+    invoke_options = { "config_file" => "test/fixtures/deploy_simple.yml", "version" => "999", "skip_hooks" => true }
+
+    Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:registry:login", [], invoke_options)
+    Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:build:deliver", [], invoke_options)
+    Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:traefik:boot", [], invoke_options)
+    Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:healthcheck:perform", [], invoke_options)
+    Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:app:stale_containers", [], invoke_options)
+    Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:app:boot", [], invoke_options)
+    Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:prune:all", [], invoke_options)
+
+    run_command("deploy", "--skip_hooks") do
+      refute_match /Running the post-deploy hook.../, output
+    end
+  end
+
+  test "deploy with missing secrets" do
+    assert_raises(KeyError) do
+      run_command("deploy", config_file: "deploy_with_secrets")
+    end
   end
 
   test "redeploy" do
-    invoke_options = { "config_file" => "test/fixtures/deploy_simple.yml", "skip_broadcast" => false, "version" => "999" }
+    invoke_options = { "config_file" => "test/fixtures/deploy_simple.yml", "version" => "999", "skip_hooks" => false }
 
     Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:build:deliver", [], invoke_options)
     Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:healthcheck:perform", [], invoke_options)
+    Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:app:stale_containers", [], invoke_options)
     Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:app:boot", [], invoke_options)
 
+    Mrsk::Commands::Hook.any_instance.stubs(:hook_exists?).returns(true)
+
+    hook_variables = { version: 999, service_version: "app@999", hosts: "1.1.1.1,1.1.1.2", command: "redeploy" }
+
     run_command("redeploy").tap do |output|
+      assert_hook_ran "pre-connect", output, **hook_variables
       assert_match /Build and push app image/, output
+      assert_hook_ran "pre-deploy", output, **hook_variables
+      assert_match /Running the pre-deploy hook.../, output
       assert_match /Ensure app can pass healthcheck/, output
+      assert_hook_ran "post-deploy", output, **hook_variables, runtime: "0"
     end
   end
 
   test "redeploy with skip_push" do
-    invoke_options = { "config_file" => "test/fixtures/deploy_simple.yml", "skip_broadcast" => false, "version" => "999" }
+    invoke_options = { "config_file" => "test/fixtures/deploy_simple.yml", "version" => "999", "skip_hooks" => false }
 
     Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:build:pull", [], invoke_options)
     Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:healthcheck:perform", [], invoke_options)
+    Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:app:stale_containers", [], invoke_options)
     Mrsk::Cli::Main.any_instance.expects(:invoke).with("mrsk:cli:app:boot", [], invoke_options)
 
     run_command("redeploy", "--skip_push").tap do |output|
@@ -120,7 +159,8 @@ class CliMainTest < CliTestCase
   end
 
   test "rollback bad version" do
-    # Mrsk::Cli::Main.any_instance.stubs(:container_available?).returns(false)
+    Thread.report_on_exception = false
+
     run_command("details") # Preheat MRSK const
 
     run_command("rollback", "nonsense").tap do |output|
@@ -130,24 +170,52 @@ class CliMainTest < CliTestCase
   end
 
   test "rollback good version" do
-    Mrsk::Cli::Main.any_instance.stubs(:container_available?).returns(true)
+    [ "web", "workers" ].each do |role|
-    SSHKit::Backend::Abstract.any_instance.expects(:capture_with_info).with(:docker, :ps, "--filter", "label=service=app", "--filter", "label=role=web", "--format", "\"{{.Names}}\"", "|", "sed 's/-/\\n/g'", "|", "tail -n 1").returns("version-to-rollback\n").at_least_once
+      SSHKit::Backend::Abstract.any_instance.expects(:capture_with_info)
-    SSHKit::Backend::Abstract.any_instance.expects(:capture_with_info).with(:docker, :ps, "--filter", "label=service=app", "--filter", "label=role=workers", "--format", "\"{{.Names}}\"", "|", "sed 's/-/\\n/g'", "|", "tail -n 1").returns("version-to-rollback\n").at_least_once
+        .with(:docker, :container, :ls, "--filter", "name=^app-#{role}-123$", "--quiet", raise_on_non_zero_exit: false)
+        .returns("").at_least_once
+      SSHKit::Backend::Abstract.any_instance.expects(:capture_with_info)
+        .with(:docker, :container, :ls, "--all", "--filter", "name=^app-#{role}-123$", "--quiet")
+        .returns("version-to-rollback\n").at_least_once
+      SSHKit::Backend::Abstract.any_instance.expects(:capture_with_info)
+        .with(:docker, :ps, "--filter", "label=service=app", "--filter", "label=role=#{role}", "--filter", "status=running", "--filter", "status=restarting", "--latest", "--format", "\"{{.Names}}\"", "|", "grep -oE \"\\-[^-]+$\"", "|", "cut -c 2-", raise_on_non_zero_exit: false)
+        .returns("version-to-rollback\n").at_least_once
+      SSHKit::Backend::Abstract.any_instance.expects(:capture_with_info)
+        .with(:docker, :container, :ls, "--all", "--filter", "name=^app-#{role}-123$", "--quiet", "|", :xargs, :docker, :inspect, "--format", "'{{if .State.Health}}{{.State.Health.Status}}{{else}}{{.State.Status}}{{end}}'")
+        .returns("running").at_least_once # health check
+    end
+
+    Mrsk::Commands::Hook.any_instance.stubs(:hook_exists?).returns(true)
+    hook_variables = { version: 123, service_version: "app@123", hosts: "1.1.1.1,1.1.1.2,1.1.1.3,1.1.1.4", command: "rollback" }
 
     run_command("rollback", "123", config_file: "deploy_with_accessories").tap do |output|
-      assert_match "Start version 123", output
+      assert_match "Start container with version 123", output
+      assert_hook_ran "pre-deploy", output, **hook_variables
+      assert_match "docker tag dhh/app:123 dhh/app:latest", output
       assert_match "docker start app-web-123", output
       assert_match "docker container ls --all --filter name=^app-web-version-to-rollback$ --quiet | xargs docker stop", output, "Should stop the container that was previously running"
+      assert_hook_ran "post-deploy", output, **hook_variables, runtime: "0"
     end
   end
 
   test "rollback without old version" do
     Mrsk::Cli::Main.any_instance.stubs(:container_available?).returns(true)
-    SSHKit::Backend::Abstract.any_instance.expects(:capture_with_info).with(:docker, :ps, "--filter", "label=service=app", "--filter", "label=role=web", "--format", "\"{{.Names}}\"", "|", "sed 's/-/\\n/g'", "|", "tail -n 1").returns("").at_least_once
+
+    Mrsk::Utils::HealthcheckPoller.stubs(:sleep)
+
+    SSHKit::Backend::Abstract.any_instance.expects(:capture_with_info)
+      .with(:docker, :container, :ls, "--filter", "name=^app-web-123$", "--quiet", raise_on_non_zero_exit: false)
+      .returns("").at_least_once
+    SSHKit::Backend::Abstract.any_instance.expects(:capture_with_info)
+      .with(:docker, :ps, "--filter", "label=service=app", "--filter", "label=role=web", "--filter", "status=running", "--filter", "status=restarting", "--latest", "--format", "\"{{.Names}}\"", "|", "grep -oE \"\\-[^-]+$\"", "|", "cut -c 2-", raise_on_non_zero_exit: false)
+      .returns("").at_least_once
+    SSHKit::Backend::Abstract.any_instance.expects(:capture_with_info)
+      .with(:docker, :container, :ls, "--all", "--filter", "name=^app-web-123$", "--quiet", "|", :xargs, :docker, :inspect, "--format", "'{{if .State.Health}}{{.State.Health.Status}}{{else}}{{.State.Status}}{{end}}'")
+      .returns("running").at_least_once # health check
 
     run_command("rollback", "123").tap do |output|
-      assert_match "Start version 123", output
+      assert_match "Start container with version 123", output
-      assert_match "docker start app-web-123", output
+      assert_match "docker start app-web-123 || docker run --detach --restart unless-stopped --name app-web-123", output
       assert_no_match "docker stop", output
     end
   end
@@ -207,9 +275,11 @@ class CliMainTest < CliTestCase
   end
 
   test "init" do
-    Pathname.any_instance.expects(:exist?).returns(false).twice
+    Pathname.any_instance.expects(:exist?).returns(false).times(3)
+    Pathname.any_instance.stubs(:mkpath)
     FileUtils.stubs(:mkdir_p)
     FileUtils.stubs(:cp_r)
+    FileUtils.stubs(:cp)
 
     run_command("init").tap do |output|
       assert_match /Created configuration file in config\/deploy.yml/, output
@@ -218,7 +288,7 @@ class CliMainTest < CliTestCase
   end
 
   test "init with existing config" do
-    Pathname.any_instance.expects(:exist?).returns(true).twice
+    Pathname.any_instance.expects(:exist?).returns(true).times(3)
 
     run_command("init").tap do |output|
       assert_match /Config file already exists in config\/deploy.yml \(remove first to create a new one\)/, output
@@ -226,9 +296,11 @@ class CliMainTest < CliTestCase
   end
 
   test "init with bundle option" do
-    Pathname.any_instance.expects(:exist?).returns(false).times(3)
+    Pathname.any_instance.expects(:exist?).returns(false).times(4)
+    Pathname.any_instance.stubs(:mkpath)
     FileUtils.stubs(:mkdir_p)
     FileUtils.stubs(:cp_r)
+    FileUtils.stubs(:cp)
 
     run_command("init", "--bundle").tap do |output|
       assert_match /Created configuration file in config\/deploy.yml/, output
@@ -241,9 +313,11 @@ class CliMainTest < CliTestCase
   end
 
   test "init with bundle option and existing binstub" do
-    Pathname.any_instance.expects(:exist?).returns(true).times(3)
+    Pathname.any_instance.expects(:exist?).returns(true).times(4)
+    Pathname.any_instance.stubs(:mkpath)
     FileUtils.stubs(:mkdir_p)
     FileUtils.stubs(:cp_r)
+    FileUtils.stubs(:cp)
 
     run_command("init", "--bundle").tap do |output|
       assert_match /Config file already exists in config\/deploy.yml \(remove first to create a new one\)/, output

test/cli/prune_test.rb

@@ -10,13 +10,14 @@ class CliPruneTest < CliTestCase
 
   test "images" do
     run_command("images").tap do |output|
-      assert_match /docker image prune --all --force --filter label=service=app --filter until=168h on 1.1.1.\d/, output
+      assert_match "docker image prune --force --filter label=service=app --filter dangling=true on 1.1.1.", output
+      assert_match "docker image ls --filter label=service=app --format '{{.ID}} {{.Repository}}:{{.Tag}}' | grep -v -w \"$(docker container ls -a --format '{{.Image}}\\|' --filter label=service=app | tr -d '\\n')dhh/app:latest\\|dhh/app:<none>\" | while read image tag; do docker rmi $tag; done on 1.1.1.", output
     end
   end
 
   test "containers" do
     run_command("containers").tap do |output|
-      assert_match /docker container prune --force --filter label=service=app --filter until=72h on 1.1.1.\d/, output
+      assert_match /docker ps -q -a --filter label=service=app --filter status=created --filter status=exited --filter status=dead | tail -n +6 | while read container_id; do docker rm $container_id; done on 1.1.1.\d/, output
     end
   end
 

test/cli/server_test.rb

@@ -1,11 +1,30 @@
 require_relative "cli_test_case"
 
 class CliServerTest < CliTestCase
-  test "bootstrap" do
+  test "bootstrap already installed" do
+    SSHKit::Backend::Abstract.any_instance.expects(:execute).with(:docker, "-v", raise_on_non_zero_exit: false).returns(true).at_least_once
+
+    assert_equal "", run_command("bootstrap")
+  end
+
+  test "bootstrap install as non-root user" do
+    SSHKit::Backend::Abstract.any_instance.expects(:execute).with(:docker, "-v", raise_on_non_zero_exit: false).returns(false).at_least_once
+    SSHKit::Backend::Abstract.any_instance.expects(:execute).with('[ "${EUID:-$(id -u)}" -eq 0 ]', raise_on_non_zero_exit: false).returns(false).at_least_once
+
+    assert_raise RuntimeError, "Docker is not installed on 1.1.1.1, 1.1.1.3, 1.1.1.4, 1.1.1.2 and can't be automatically installed without having root access and the `curl` command available. Install Docker manually: https://docs.docker.com/engine/install/" do
+      run_command("bootstrap")
+    end
+  end
+
+  test "bootstrap install as root user" do
+    SSHKit::Backend::Abstract.any_instance.expects(:execute).with(:docker, "-v", raise_on_non_zero_exit: false).returns(false).at_least_once
+    SSHKit::Backend::Abstract.any_instance.expects(:execute).with('[ "${EUID:-$(id -u)}" -eq 0 ]', raise_on_non_zero_exit: false).returns(true).at_least_once
+    SSHKit::Backend::Abstract.any_instance.expects(:execute).with(:curl, "-fsSL", "https://get.docker.com", "|", :sh).at_least_once
+
     run_command("bootstrap").tap do |output|
-      assert_match /which curl/, output
+      ("1.1.1.1".."1.1.1.4").map do |host|
-      assert_match /which docker/, output
+        assert_match "Missing Docker on #{host}. Installing…", output
-      assert_match /apt-get update -y && apt-get install curl docker.io -y/, output
+      end
     end
   end
 

test/cli/traefik_test.rb

@@ -4,16 +4,24 @@ class CliTraefikTest < CliTestCase
   test "boot" do
     run_command("boot").tap do |output|
       assert_match "docker login", output
-      assert_match "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" #{Mrsk::Commands::Traefik::DEFAULT_IMAGE} --providers.docker --log.level=DEBUG", output
+      assert_match "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" #{Mrsk::Commands::Traefik::DEFAULT_IMAGE} --providers.docker --log.level=\"DEBUG\"", output
     end
   end
 
   test "reboot" do
-    Mrsk::Cli::Traefik.any_instance.expects(:stop)
+    Mrsk::Commands::Registry.any_instance.expects(:login).twice
-    Mrsk::Cli::Traefik.any_instance.expects(:remove_container)
-    Mrsk::Cli::Traefik.any_instance.expects(:boot)
 
-    run_command("reboot")
+    run_command("reboot").tap do |output|
+      assert_match "docker container stop traefik", output
+      assert_match "docker container prune --force --filter label=org.opencontainers.image.title=Traefik", output
+      assert_match "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" #{Mrsk::Commands::Traefik::DEFAULT_IMAGE} --providers.docker --log.level=\"DEBUG\"", output
+    end
+  end
+
+  test "reboot --rolling" do
+    run_command("reboot", "--rolling").tap do |output|
+      assert_match "Running docker container prune --force --filter label=org.opencontainers.image.title=Traefik on 1.1.1.1", output.lines[3]
+    end
   end
 
   test "start" do

test/commander_test.rb

@@ -2,9 +2,7 @@ require "test_helper"
 
 class CommanderTest < ActiveSupport::TestCase
   setup do
-    @mrsk = Mrsk::Commander.new.tap do |mrsk|
+    configure_with(:deploy_with_roles)
-      mrsk.configure config_file: Pathname.new(File.expand_path("fixtures/deploy_with_roles.yml", __dir__))
-    end
   end
 
   test "lazy configuration" do
@@ -47,12 +45,35 @@ class CommanderTest < ActiveSupport::TestCase
   end
 
   test "primary_host with specific hosts via role" do
-    @mrsk.specific_roles = "web"
+    @mrsk.specific_roles = "workers"
-    assert_equal "1.1.1.1", @mrsk.primary_host
+    assert_equal "1.1.1.3", @mrsk.primary_host
   end
 
   test "roles_on" do
     assert_equal [ "web" ], @mrsk.roles_on("1.1.1.1")
     assert_equal [ "workers" ], @mrsk.roles_on("1.1.1.3")
   end
+
+  test "default group strategy" do
+    assert_empty @mrsk.boot_strategy
+  end
+
+  test "specific limit group strategy" do
+    configure_with(:deploy_with_boot_strategy)
+
+    assert_equal({ in: :groups, limit: 3, wait: 2 }, @mrsk.boot_strategy)
+  end
+
+  test "percentage-based group strategy" do
+    configure_with(:deploy_with_percentage_boot_strategy)
+
+    assert_equal({ in: :groups, limit: 1, wait: 2 }, @mrsk.boot_strategy)
+  end
+
+  private
+    def configure_with(variant)
+      @mrsk = Mrsk::Commander.new.tap do |mrsk|
+        mrsk.configure config_file: Pathname.new(File.expand_path("fixtures/#{variant}.yml", __dir__))
+      end
+    end
 end

test/commands/app_test.rb

@@ -13,15 +13,21 @@ class CommandsAppTest < ActiveSupport::TestCase
 
   test "run" do
     assert_equal \
-      "docker run --detach --restart unless-stopped --name app-web-999 -e MRSK_CONTAINER_NAME=\"app-web-999\" -e RAILS_MASTER_KEY=\"456\" --log-opt max-size=\"10m\" --label service=\"app\" --label role=\"web\" --label traefik.http.routers.app-web.rule=\"PathPrefix(\\`/\\`)\" --label traefik.http.services.app-web.loadbalancer.healthcheck.path=\"/up\" --label traefik.http.services.app-web.loadbalancer.healthcheck.interval=\"1s\" --label traefik.http.middlewares.app-web-retry.retry.attempts=\"5\" --label traefik.http.middlewares.app-web-retry.retry.initialinterval=\"500ms\" --label traefik.http.routers.app-web.middlewares=\"app-web-retry@docker\" dhh/app:999",
+      "docker run --detach --restart unless-stopped --name app-web-999 -e MRSK_CONTAINER_NAME=\"app-web-999\" -e RAILS_MASTER_KEY=\"456\" --health-cmd \"curl -f http://localhost:3000/up || exit 1\" --health-interval \"1s\" --log-opt max-size=\"10m\" --label service=\"app\" --label role=\"web\" --label traefik.http.services.app-web.loadbalancer.server.scheme=\"http\" --label traefik.http.routers.app-web.rule=\"PathPrefix(\\`/\\`)\" --label traefik.http.middlewares.app-web-retry.retry.attempts=\"5\" --label traefik.http.middlewares.app-web-retry.retry.initialinterval=\"500ms\" --label traefik.http.routers.app-web.middlewares=\"app-web-retry@docker\" dhh/app:999",
       new_command.run.join(" ")
   end
 
+  test "run with hostname" do
+    assert_equal \
+      "docker run --detach --restart unless-stopped --name app-web-999 --hostname myhost -e MRSK_CONTAINER_NAME=\"app-web-999\" -e RAILS_MASTER_KEY=\"456\" --health-cmd \"curl -f http://localhost:3000/up || exit 1\" --health-interval \"1s\" --log-opt max-size=\"10m\" --label service=\"app\" --label role=\"web\" --label traefik.http.services.app-web.loadbalancer.server.scheme=\"http\" --label traefik.http.routers.app-web.rule=\"PathPrefix(\\`/\\`)\" --label traefik.http.middlewares.app-web-retry.retry.attempts=\"5\" --label traefik.http.middlewares.app-web-retry.retry.initialinterval=\"500ms\" --label traefik.http.routers.app-web.middlewares=\"app-web-retry@docker\" dhh/app:999",
+      new_command.run(hostname: "myhost").join(" ")
+  end
+
   test "run with volumes" do
     @config[:volumes] = ["/local/path:/container/path" ]
 
     assert_equal \
-      "docker run --detach --restart unless-stopped --name app-web-999 -e MRSK_CONTAINER_NAME=\"app-web-999\" -e RAILS_MASTER_KEY=\"456\" --log-opt max-size=\"10m\" --volume /local/path:/container/path --label service=\"app\" --label role=\"web\" --label traefik.http.routers.app-web.rule=\"PathPrefix(\\`/\\`)\" --label traefik.http.services.app-web.loadbalancer.healthcheck.path=\"/up\" --label traefik.http.services.app-web.loadbalancer.healthcheck.interval=\"1s\" --label traefik.http.middlewares.app-web-retry.retry.attempts=\"5\" --label traefik.http.middlewares.app-web-retry.retry.initialinterval=\"500ms\" --label traefik.http.routers.app-web.middlewares=\"app-web-retry@docker\" dhh/app:999",
+      "docker run --detach --restart unless-stopped --name app-web-999 -e MRSK_CONTAINER_NAME=\"app-web-999\" -e RAILS_MASTER_KEY=\"456\" --health-cmd \"curl -f http://localhost:3000/up || exit 1\" --health-interval \"1s\" --log-opt max-size=\"10m\" --volume /local/path:/container/path --label service=\"app\" --label role=\"web\" --label traefik.http.services.app-web.loadbalancer.server.scheme=\"http\" --label traefik.http.routers.app-web.rule=\"PathPrefix(\\`/\\`)\" --label traefik.http.middlewares.app-web-retry.retry.attempts=\"5\" --label traefik.http.middlewares.app-web-retry.retry.initialinterval=\"500ms\" --label traefik.http.routers.app-web.middlewares=\"app-web-retry@docker\" dhh/app:999",
       new_command.run.join(" ")
   end
 
@@ -29,7 +35,23 @@ class CommandsAppTest < ActiveSupport::TestCase
     @config[:healthcheck] = { "path" => "/healthz" }
 
     assert_equal \
-      "docker run --detach --restart unless-stopped --name app-web-999 -e MRSK_CONTAINER_NAME=\"app-web-999\" -e RAILS_MASTER_KEY=\"456\" --log-opt max-size=\"10m\" --label service=\"app\" --label role=\"web\" --label traefik.http.routers.app-web.rule=\"PathPrefix(\\`/\\`)\" --label traefik.http.services.app-web.loadbalancer.healthcheck.path=\"/healthz\" --label traefik.http.services.app-web.loadbalancer.healthcheck.interval=\"1s\" --label traefik.http.middlewares.app-web-retry.retry.attempts=\"5\" --label traefik.http.middlewares.app-web-retry.retry.initialinterval=\"500ms\" --label traefik.http.routers.app-web.middlewares=\"app-web-retry@docker\" dhh/app:999",
+      "docker run --detach --restart unless-stopped --name app-web-999 -e MRSK_CONTAINER_NAME=\"app-web-999\" -e RAILS_MASTER_KEY=\"456\" --health-cmd \"curl -f http://localhost:3000/healthz || exit 1\" --health-interval \"1s\" --log-opt max-size=\"10m\" --label service=\"app\" --label role=\"web\" --label traefik.http.services.app-web.loadbalancer.server.scheme=\"http\" --label traefik.http.routers.app-web.rule=\"PathPrefix(\\`/\\`)\" --label traefik.http.middlewares.app-web-retry.retry.attempts=\"5\" --label traefik.http.middlewares.app-web-retry.retry.initialinterval=\"500ms\" --label traefik.http.routers.app-web.middlewares=\"app-web-retry@docker\" dhh/app:999",
+      new_command.run.join(" ")
+  end
+
+  test "run with custom healthcheck command" do
+    @config[:healthcheck] = { "cmd" => "/bin/up" }
+
+    assert_equal \
+      "docker run --detach --restart unless-stopped --name app-web-999 -e MRSK_CONTAINER_NAME=\"app-web-999\" -e RAILS_MASTER_KEY=\"456\" --health-cmd \"/bin/up\" --health-interval \"1s\" --log-opt max-size=\"10m\" --label service=\"app\" --label role=\"web\" --label traefik.http.services.app-web.loadbalancer.server.scheme=\"http\" --label traefik.http.routers.app-web.rule=\"PathPrefix(\\`/\\`)\" --label traefik.http.middlewares.app-web-retry.retry.attempts=\"5\" --label traefik.http.middlewares.app-web-retry.retry.initialinterval=\"500ms\" --label traefik.http.routers.app-web.middlewares=\"app-web-retry@docker\" dhh/app:999",
+      new_command.run.join(" ")
+  end
+
+  test "run with role-specific healthcheck options" do
+    @config[:servers] = { "web" => { "hosts" => [ "1.1.1.1" ], "healthcheck" => { "cmd" => "/bin/healthy" } } }
+
+    assert_equal \
+      "docker run --detach --restart unless-stopped --name app-web-999 -e MRSK_CONTAINER_NAME=\"app-web-999\" -e RAILS_MASTER_KEY=\"456\" --health-cmd \"/bin/healthy\" --health-interval \"1s\" --log-opt max-size=\"10m\" --label service=\"app\" --label role=\"web\" --label traefik.http.services.app-web.loadbalancer.server.scheme=\"http\" --label traefik.http.routers.app-web.rule=\"PathPrefix(\\`/\\`)\" --label traefik.http.middlewares.app-web-retry.retry.attempts=\"5\" --label traefik.http.middlewares.app-web-retry.retry.initialinterval=\"500ms\" --label traefik.http.routers.app-web.middlewares=\"app-web-retry@docker\" dhh/app:999",
       new_command.run.join(" ")
   end
 
@@ -44,7 +66,7 @@ class CommandsAppTest < ActiveSupport::TestCase
     @config[:logging] = { "driver" => "local", "options" => { "max-size" => "100m", "max-file" => "3" } }
 
     assert_equal \
-      "docker run --detach --restart unless-stopped --name app-web-999 -e MRSK_CONTAINER_NAME=\"app-web-999\" -e RAILS_MASTER_KEY=\"456\" --log-driver \"local\" --log-opt max-size=\"100m\" --log-opt max-file=\"3\" --label service=\"app\" --label role=\"web\" --label traefik.http.routers.app-web.rule=\"PathPrefix(\\`/\\`)\" --label traefik.http.services.app-web.loadbalancer.healthcheck.path=\"/up\" --label traefik.http.services.app-web.loadbalancer.healthcheck.interval=\"1s\" --label traefik.http.middlewares.app-web-retry.retry.attempts=\"5\" --label traefik.http.middlewares.app-web-retry.retry.initialinterval=\"500ms\" --label traefik.http.routers.app-web.middlewares=\"app-web-retry@docker\" dhh/app:999",
+      "docker run --detach --restart unless-stopped --name app-web-999 -e MRSK_CONTAINER_NAME=\"app-web-999\" -e RAILS_MASTER_KEY=\"456\" --health-cmd \"curl -f http://localhost:3000/up || exit 1\" --health-interval \"1s\" --log-driver \"local\" --log-opt max-size=\"100m\" --log-opt max-file=\"3\" --label service=\"app\" --label role=\"web\" --label traefik.http.services.app-web.loadbalancer.server.scheme=\"http\" --label traefik.http.routers.app-web.rule=\"PathPrefix(\\`/\\`)\" --label traefik.http.middlewares.app-web-retry.retry.attempts=\"5\" --label traefik.http.middlewares.app-web-retry.retry.initialinterval=\"500ms\" --label traefik.http.routers.app-web.middlewares=\"app-web-retry@docker\" dhh/app:999",
       new_command.run.join(" ")
   end
 
@@ -61,16 +83,28 @@ class CommandsAppTest < ActiveSupport::TestCase
       new_command.start.join(" ")
   end
 
+  test "start_or_run" do
+    assert_equal \
+      "docker start app-web-999 || docker run --detach --restart unless-stopped --name app-web-999 -e MRSK_CONTAINER_NAME=\"app-web-999\" -e RAILS_MASTER_KEY=\"456\" --health-cmd \"curl -f http://localhost:3000/up || exit 1\" --health-interval \"1s\" --log-opt max-size=\"10m\" --label service=\"app\" --label role=\"web\" --label traefik.http.services.app-web.loadbalancer.server.scheme=\"http\" --label traefik.http.routers.app-web.rule=\"PathPrefix(\\`/\\`)\" --label traefik.http.middlewares.app-web-retry.retry.attempts=\"5\" --label traefik.http.middlewares.app-web-retry.retry.initialinterval=\"500ms\" --label traefik.http.routers.app-web.middlewares=\"app-web-retry@docker\" dhh/app:999",
+      new_command.start_or_run.join(" ")
+  end
+
+  test "start_or_run with hostname" do
+    assert_equal \
+      "docker start app-web-999 || docker run --detach --restart unless-stopped --name app-web-999 --hostname myhost -e MRSK_CONTAINER_NAME=\"app-web-999\" -e RAILS_MASTER_KEY=\"456\" --health-cmd \"curl -f http://localhost:3000/up || exit 1\" --health-interval \"1s\" --log-opt max-size=\"10m\" --label service=\"app\" --label role=\"web\" --label traefik.http.services.app-web.loadbalancer.server.scheme=\"http\" --label traefik.http.routers.app-web.rule=\"PathPrefix(\\`/\\`)\" --label traefik.http.middlewares.app-web-retry.retry.attempts=\"5\" --label traefik.http.middlewares.app-web-retry.retry.initialinterval=\"500ms\" --label traefik.http.routers.app-web.middlewares=\"app-web-retry@docker\" dhh/app:999",
+      new_command.start_or_run(hostname: "myhost").join(" ")
+  end
+
   test "stop" do
     assert_equal \
-      "docker ps --quiet --filter label=service=app --filter label=role=web | xargs docker stop",
+      "docker ps --quiet --filter label=service=app --filter label=role=web --filter status=running --filter status=restarting --latest | xargs docker stop",
       new_command.stop.join(" ")
   end
 
   test "stop with custom stop wait time" do
     @config[:stop_wait_time] = 30
     assert_equal \
-      "docker ps --quiet --filter label=service=app --filter label=role=web | xargs docker stop -t 30",
+      "docker ps --quiet --filter label=service=app --filter label=role=web --filter status=running --filter status=restarting --latest | xargs docker stop -t 30",
       new_command.stop.join(" ")
   end
 
@@ -96,37 +130,37 @@ class CommandsAppTest < ActiveSupport::TestCase
 
   test "logs" do
     assert_equal \
-      "docker ps --quiet --filter label=service=app --filter label=role=web | xargs docker logs 2>&1",
+      "docker ps --quiet --filter label=service=app --filter label=role=web --filter status=running --filter status=restarting --latest | xargs docker logs 2>&1",
       new_command.logs.join(" ")
 
     assert_equal \
-      "docker ps --quiet --filter label=service=app --filter label=role=web | xargs docker logs --since 5m 2>&1",
+      "docker ps --quiet --filter label=service=app --filter label=role=web --filter status=running --filter status=restarting --latest | xargs docker logs --since 5m 2>&1",
       new_command.logs(since: "5m").join(" ")
 
     assert_equal \
-      "docker ps --quiet --filter label=service=app --filter label=role=web | xargs docker logs --tail 100 2>&1",
+      "docker ps --quiet --filter label=service=app --filter label=role=web --filter status=running --filter status=restarting --latest | xargs docker logs --tail 100 2>&1",
       new_command.logs(lines: "100").join(" ")
 
     assert_equal \
-      "docker ps --quiet --filter label=service=app --filter label=role=web | xargs docker logs --since 5m --tail 100 2>&1",
+      "docker ps --quiet --filter label=service=app --filter label=role=web --filter status=running --filter status=restarting --latest | xargs docker logs --since 5m --tail 100 2>&1",
       new_command.logs(since: "5m", lines: "100").join(" ")
 
     assert_equal \
-      "docker ps --quiet --filter label=service=app --filter label=role=web | xargs docker logs 2>&1 | grep 'my-id'",
+      "docker ps --quiet --filter label=service=app --filter label=role=web --filter status=running --filter status=restarting --latest | xargs docker logs 2>&1 | grep 'my-id'",
       new_command.logs(grep: "my-id").join(" ")
 
     assert_equal \
-      "docker ps --quiet --filter label=service=app --filter label=role=web | xargs docker logs --since 5m 2>&1 | grep 'my-id'",
+      "docker ps --quiet --filter label=service=app --filter label=role=web --filter status=running --filter status=restarting --latest | xargs docker logs --since 5m 2>&1 | grep 'my-id'",
       new_command.logs(since: "5m", grep: "my-id").join(" ")
   end
 
   test "follow logs" do
     assert_match \
-      "docker ps --quiet --filter label=service=app --filter label=role=web | xargs docker logs --timestamps --tail 10 --follow 2>&1",
+      "docker ps --quiet --filter label=service=app --filter label=role=web --filter status=running --filter status=restarting --latest | xargs docker logs --timestamps --tail 10 --follow 2>&1",
       new_command.follow_logs(host: "app-1")
 
     assert_match \
-      "docker ps --quiet --filter label=service=app --filter label=role=web | xargs docker logs --timestamps --tail 10 --follow 2>&1 | grep \"Completed\"",
+      "docker ps --quiet --filter label=service=app --filter label=role=web --filter status=running --filter status=restarting --latest | xargs docker logs --timestamps --tail 10 --follow 2>&1 | grep \"Completed\"",
       new_command.follow_logs(host: "app-1", grep: "Completed")
   end
 
@@ -177,18 +211,22 @@ class CommandsAppTest < ActiveSupport::TestCase
     assert_equal "ssh -J root@2.2.2.2 -t app@1.1.1.1 'ls'", new_command.run_over_ssh("ls", host: "1.1.1.1")
   end
 
-
+  test "run over ssh with proxy_command" do
-  test "current_container_id" do
+    @config[:ssh] = { "proxy_command" => "ssh -W %h:%p user@proxy-server" }
-    assert_equal \
+    assert_equal "ssh -o ProxyCommand='ssh -W %h:%p user@proxy-server' -t root@1.1.1.1 'ls'", new_command.run_over_ssh("ls", host: "1.1.1.1")
-      "docker ps --quiet --filter label=service=app --filter label=role=web",
-      new_command.current_container_id.join(" ")
   end
 
-  test "current_container_id with destination" do
+  test "current_running_container_id" do
+    assert_equal \
+      "docker ps --quiet --filter label=service=app --filter label=role=web --filter status=running --filter status=restarting --latest",
+      new_command.current_running_container_id.join(" ")
+  end
+
+  test "current_running_container_id with destination" do
     @destination = "staging"
     assert_equal \
-      "docker ps --quiet --filter label=service=app --filter label=destination=staging --filter label=role=web",
+      "docker ps --quiet --filter label=service=app --filter label=destination=staging --filter label=role=web --filter status=running --filter status=restarting --latest",
-      new_command.current_container_id.join(" ")
+      new_command.current_running_container_id.join(" ")
   end
 
   test "container_id_for" do
@@ -199,10 +237,20 @@ class CommandsAppTest < ActiveSupport::TestCase
 
   test "current_running_version" do
     assert_equal \
-      "docker ps --filter label=service=app --filter label=role=web --format \"{{.Names}}\" | sed 's/-/\\n/g' | tail -n 1",
+      "docker ps --filter label=service=app --filter label=role=web --filter status=running --filter status=restarting --latest --format \"{{.Names}}\" | grep -oE \"\\-[^-]+$\" | cut -c 2-",
       new_command.current_running_version.join(" ")
   end
 
+  test "list_versions" do
+    assert_equal \
+      "docker ps --filter label=service=app --filter label=role=web --format \"{{.Names}}\" | grep -oE \"\\-[^-]+$\" | cut -c 2-",
+      new_command.list_versions.join(" ")
+
+    assert_equal \
+      "docker ps --filter label=service=app --filter label=role=web --filter status=running --filter status=restarting --latest --format \"{{.Names}}\" | grep -oE \"\\-[^-]+$\" | cut -c 2-",
+      new_command.list_versions("--latest", statuses: [ :running, :restarting ]).join(" ")
+  end
+
   test "list_containers" do
     assert_equal \
       "docker container ls --all --filter label=service=app --filter label=role=web",
@@ -267,6 +315,12 @@ class CommandsAppTest < ActiveSupport::TestCase
       new_command.remove_images.join(" ")
   end
 
+  test "tag_current_as_latest" do
+    assert_equal \
+      "docker tag dhh/app:999 dhh/app:latest",
+      new_command.tag_current_as_latest.join(" ")
+  end
+
   private
     def new_command(role: "web")
       Mrsk::Commands::App.new(Mrsk::Configuration.new(@config, destination: @destination, version: "999"), role: role)

test/commands/auditor_test.rb

@@ -1,43 +1,64 @@
 require "test_helper"
+require "active_support/testing/time_helpers"
 
 class CommandsAuditorTest < ActiveSupport::TestCase
+  include ActiveSupport::Testing::TimeHelpers
+
   setup do
+    freeze_time
+
     @config = {
-      service: "app", image: "dhh/app", registry: { "username" => "dhh", "password" => "secret" }, servers: [ "1.1.1.1" ],
+      service: "app", image: "dhh/app", registry: { "username" => "dhh", "password" => "secret" }, servers: [ "1.1.1.1" ]
-      audit_broadcast_cmd: "bin/audit_broadcast"
     }
+
+    @auditor = new_command
+    @performer = `whoami`.strip
+    @recorded_at = Time.now.utc.iso8601
   end
 
   test "record" do
-    assert_match \
+    assert_equal [
-      /echo '.* app removed container' >> mrsk-app-audit.log/,
+      :echo,
-      new_command.record("app removed container").join(" ")
+      "[#{@recorded_at}] [#{@performer}]",
+      "app removed container",
+      ">>", "mrsk-app-audit.log"
+    ], @auditor.record("app removed container")
   end
 
   test "record with destination" do
-    @destination = "staging"
+    new_command(destination: "staging").tap do |auditor|
-
+      assert_equal [
-    assert_match \
+        :echo,
-      /echo '.* app removed container' >> mrsk-app-staging-audit.log/,
+        "[#{@recorded_at}] [#{@performer}] [staging]",
-      new_command.record("app removed container").join(" ")
+        "app removed container",
+        ">>", "mrsk-app-staging-audit.log"
+      ], auditor.record("app removed container")
+    end
   end
 
-  test "record with role" do
+  test "record with command details" do
-    @role = "web"
+    new_command(role: "web").tap do |auditor|
-
+      assert_equal [
-    assert_match \
+        :echo,
-      /echo '.* \[web\] app removed container' >> mrsk-app-audit.log/,
+        "[#{@recorded_at}] [#{@performer}] [web]",
-      new_command.record("app removed container").join(" ")
+        "app removed container",
+        ">>", "mrsk-app-audit.log"
+      ], auditor.record("app removed container")
+    end
   end
 
-  test "broadcast" do
+  test "record with arg details" do
-    assert_match \
+    assert_equal [
-      /bin\/audit_broadcast '\[.*\] app removed container'/,
+      :echo,
-      new_command.broadcast("app removed container").join(" ")
+      "[#{@recorded_at}] [#{@performer}] [value]",
+      "app removed container",
+      ">>", "mrsk-app-audit.log"
+    ], @auditor.record("app removed container", detail: "value")
   end
 
+
   private
-    def new_command
+    def new_command(destination: nil, **details)
-      Mrsk::Commands::Auditor.new(Mrsk::Configuration.new(@config, destination: @destination, version: "123"), role: @role)
+      Mrsk::Commands::Auditor.new(Mrsk::Configuration.new(@config, destination: destination, version: "123"), **details)
     end
 end

test/commands/builder_test.rb

@@ -6,10 +6,10 @@ class CommandsBuilderTest < ActiveSupport::TestCase
   end
 
   test "target multiarch by default" do
-    builder = new_builder_command
+    builder = new_builder_command(builder: { "cache" => { "type" => "gha" }})
     assert_equal "multiarch", builder.name
     assert_equal \
-      "docker buildx build --push --platform linux/amd64,linux/arm64 --builder mrsk-app-multiarch -t dhh/app:123 -t dhh/app:latest --label service=\"app\" --file Dockerfile .",
+      "docker buildx build --push --platform linux/amd64,linux/arm64 --builder mrsk-app-multiarch -t dhh/app:123 -t dhh/app:latest --cache-to type=gha --cache-from type=gha --label service=\"app\" --file Dockerfile .",
       builder.push.join(" ")
   end
 
@@ -21,19 +21,27 @@ class CommandsBuilderTest < ActiveSupport::TestCase
       builder.push.join(" ")
   end
 
+  test "target native cached when multiarch is off and cache is set" do
+    builder = new_builder_command(builder: { "multiarch" => false, "cache" => { "type" => "gha" }})
+    assert_equal "native/cached", builder.name
+    assert_equal \
+      "docker buildx build --push -t dhh/app:123 -t dhh/app:latest --cache-to type=gha --cache-from type=gha --label service=\"app\" --file Dockerfile .",
+      builder.push.join(" ")
+  end
+
   test "target multiarch remote when local and remote is set" do
-    builder = new_builder_command(builder: { "local" => { }, "remote" => { } })
+    builder = new_builder_command(builder: { "local" => { }, "remote" => { }, "cache" => { "type" => "gha" } })
     assert_equal "multiarch/remote", builder.name
     assert_equal \
-      "docker buildx build --push --platform linux/amd64,linux/arm64 --builder mrsk-app-multiarch-remote -t dhh/app:123 -t dhh/app:latest --label service=\"app\" --file Dockerfile .",
+      "docker buildx build --push --platform linux/amd64,linux/arm64 --builder mrsk-app-multiarch-remote -t dhh/app:123 -t dhh/app:latest --cache-to type=gha --cache-from type=gha --label service=\"app\" --file Dockerfile .",
       builder.push.join(" ")
   end
 
   test "target native remote when only remote is set" do
-    builder = new_builder_command(builder: { "remote" => { "arch" => "amd64" } })
+    builder = new_builder_command(builder: { "remote" => { "arch" => "amd64" }, "cache" => { "type" => "gha" } })
     assert_equal "native/remote", builder.name
     assert_equal \
-      "docker buildx build --push --platform linux/amd64 --builder mrsk-app-native-remote -t dhh/app:123 -t dhh/app:latest --label service=\"app\" --file Dockerfile .",
+      "docker buildx build --push --platform linux/amd64 --builder mrsk-app-native-remote -t dhh/app:123 -t dhh/app:latest --cache-to type=gha --cache-from type=gha --label service=\"app\" --file Dockerfile .",
       builder.push.join(" ")
   end
 
@@ -52,12 +60,21 @@ class CommandsBuilderTest < ActiveSupport::TestCase
   end
 
   test "build dockerfile" do
+    Pathname.any_instance.expects(:exist?).returns(true).once
     builder = new_builder_command(builder: { "dockerfile" => "Dockerfile.xyz" })
     assert_equal \
       "-t dhh/app:123 -t dhh/app:latest --label service=\"app\" --file Dockerfile.xyz",
       builder.target.build_options.join(" ")
   end
 
+  test "missing dockerfile" do
+    Pathname.any_instance.expects(:exist?).returns(false).once
+    builder = new_builder_command(builder: { "dockerfile" => "Dockerfile.xyz" })
+    assert_raises(Mrsk::Commands::Builder::Base::BuilderError) do
+      builder.target.build_options.join(" ")
+    end
+  end
+
   test "build context" do
     builder = new_builder_command(builder: { "context" => ".." })
     assert_equal \
@@ -79,7 +96,7 @@ class CommandsBuilderTest < ActiveSupport::TestCase
       builder.push.join(" ")
   end
 
-  test "native push with with build secrets" do
+  test "native push with build secrets" do
     builder = new_builder_command(builder: { "multiarch" => false, "secrets" => [ "a", "b" ] })
     assert_equal \
       "docker build -t dhh/app:123 -t dhh/app:latest --label service=\"app\" --secret id=\"a\" --secret id=\"b\" --file Dockerfile . && docker push dhh/app:123 && docker push dhh/app:latest",

test/commands/docker_test.rb

@@ -0,0 +1,26 @@
+require "test_helper"
+
+class CommandsDockerTest < ActiveSupport::TestCase
+  setup do
+    @config = {
+      service: "app", image: "dhh/app", registry: { "username" => "dhh", "password" => "secret" }, servers: [ "1.1.1.1" ]
+    }
+    @docker = Mrsk::Commands::Docker.new(Mrsk::Configuration.new(@config))
+  end
+
+  test "install" do
+    assert_equal "curl -fsSL https://get.docker.com | sh", @docker.install.join(" ")
+  end
+
+  test "installed?" do
+    assert_equal "docker -v", @docker.installed?.join(" ")
+  end
+
+  test "running?" do
+    assert_equal "docker version", @docker.running?.join(" ")
+  end
+
+  test "superuser?" do
+    assert_equal '[ "${EUID:-$(id -u)}" -eq 0 ]', @docker.superuser?.join(" ")
+  end
+end

test/commands/healthcheck_test.rb

@@ -10,7 +10,7 @@ class CommandsHealthcheckTest < ActiveSupport::TestCase
 
   test "run" do
     assert_equal \
-      "docker run --detach --name healthcheck-app-123 --publish 3999:3000 --label service=healthcheck-app -e MRSK_CONTAINER_NAME=\"healthcheck-app\" dhh/app:123",
+      "docker run --detach --name healthcheck-app-123 --publish 3999:3000 --label service=healthcheck-app -e MRSK_CONTAINER_NAME=\"healthcheck-app\" --health-cmd \"curl -f http://localhost:3000/up || exit 1\" --health-interval \"1s\" dhh/app:123",
       new_command.run.join(" ")
   end
 
@@ -18,7 +18,7 @@ class CommandsHealthcheckTest < ActiveSupport::TestCase
     @config[:healthcheck] = { "port" => 3001 }
 
     assert_equal \
-      "docker run --detach --name healthcheck-app-123 --publish 3999:3001 --label service=healthcheck-app -e MRSK_CONTAINER_NAME=\"healthcheck-app\" dhh/app:123",
+      "docker run --detach --name healthcheck-app-123 --publish 3999:3001 --label service=healthcheck-app -e MRSK_CONTAINER_NAME=\"healthcheck-app\" --health-cmd \"curl -f http://localhost:3001/up || exit 1\" --health-interval \"1s\" dhh/app:123",
       new_command.run.join(" ")
   end
 
@@ -26,29 +26,35 @@ class CommandsHealthcheckTest < ActiveSupport::TestCase
     @destination = "staging"
 
     assert_equal \
-      "docker run --detach --name healthcheck-app-staging-123 --publish 3999:3000 --label service=healthcheck-app-staging -e MRSK_CONTAINER_NAME=\"healthcheck-app-staging\" dhh/app:123",
+      "docker run --detach --name healthcheck-app-staging-123 --publish 3999:3000 --label service=healthcheck-app-staging -e MRSK_CONTAINER_NAME=\"healthcheck-app-staging\" --health-cmd \"curl -f http://localhost:3000/up || exit 1\" --health-interval \"1s\" dhh/app:123",
+      new_command.run.join(" ")
+  end
+
+  test "run with custom healthcheck" do
+    @config[:healthcheck] = { "cmd" => "/bin/up" }
+
+    assert_equal \
+      "docker run --detach --name healthcheck-app-123 --publish 3999:3000 --label service=healthcheck-app -e MRSK_CONTAINER_NAME=\"healthcheck-app\" --health-cmd \"/bin/up\" --health-interval \"1s\" dhh/app:123",
       new_command.run.join(" ")
   end
 
   test "run with custom options" do
     @config[:servers] = { "web" => { "hosts" => [ "1.1.1.1" ], "options" => { "mount" => "somewhere" } } }
     assert_equal \
-      "docker run --detach --name healthcheck-app-123 --publish 3999:3000 --label service=healthcheck-app -e MRSK_CONTAINER_NAME=\"healthcheck-app\" --mount \"somewhere\" dhh/app:123",
+      "docker run --detach --name healthcheck-app-123 --publish 3999:3000 --label service=healthcheck-app -e MRSK_CONTAINER_NAME=\"healthcheck-app\" --health-cmd \"curl -f http://localhost:3000/up || exit 1\" --health-interval \"1s\" --mount \"somewhere\" dhh/app:123",
       new_command.run.join(" ")
   end
 
-  test "curl" do
+  test "status" do
     assert_equal \
-      "curl --silent --output /dev/null --write-out '%{http_code}' --max-time 2 http://localhost:3999/up",
+      "docker container ls --all --filter name=^healthcheck-app-123$ --quiet | xargs docker inspect --format '{{if .State.Health}}{{.State.Health.Status}}{{else}}{{.State.Status}}{{end}}'",
-      new_command.curl.join(" ")
+      new_command.status.join(" ")
   end
 
-  test "curl with custom path" do
+  test "container_health_log" do
-    @config[:healthcheck] = { "path" => "/healthz" }
-
     assert_equal \
-      "curl --silent --output /dev/null --write-out '%{http_code}' --max-time 2 http://localhost:3999/healthz",
+      "docker container ls --all --filter name=^healthcheck-app-123$ --quiet | xargs docker inspect --format '{{json .State.Health}}'",
-      new_command.curl.join(" ")
+      new_command.container_health_log.join(" ")
   end
 
   test "stop" do

test/commands/hook_test.rb

@@ -0,0 +1,44 @@
+require "test_helper"
+
+class CommandsHookTest < ActiveSupport::TestCase
+  include ActiveSupport::Testing::TimeHelpers
+
+  setup do
+    freeze_time
+
+    @config = {
+      service: "app", image: "dhh/app", registry: { "username" => "dhh", "password" => "secret" }, servers: [ "1.1.1.1" ],
+      traefik: { "args" => { "accesslog.format" => "json", "metrics.prometheus.buckets" => "0.1,0.3,1.2,5.0" } }
+    }
+
+    @performer = `whoami`.strip
+    @recorded_at = Time.now.utc.iso8601
+  end
+
+  test "run" do
+    assert_equal [
+      ".mrsk/hooks/foo",
+      { env: {
+        "MRSK_RECORDED_AT" => @recorded_at,
+        "MRSK_PERFORMER" => @performer,
+        "MRSK_VERSION" => "123",
+        "MRSK_SERVICE_VERSION" => "app@123" } }
+    ], new_command.run("foo")
+  end
+
+  test "run with custom hooks_path" do
+    assert_equal [
+      "custom/hooks/path/foo",
+      { env: {
+        "MRSK_RECORDED_AT" => @recorded_at,
+        "MRSK_PERFORMER" => @performer,
+        "MRSK_VERSION" => "123",
+        "MRSK_SERVICE_VERSION" => "app@123" } }
+    ], new_command(hooks_path: "custom/hooks/path").run("foo")
+  end
+
+  private
+    def new_command(**extra_config)
+      Mrsk::Commands::Hook.new(Mrsk::Configuration.new(@config.merge(**extra_config), version: "123"))
+    end
+end

test/commands/lock_test.rb

@@ -10,19 +10,19 @@ class CommandsLockTest < ActiveSupport::TestCase
 
   test "status" do
     assert_equal \
-      "stat mrsk_lock > /dev/null && cat mrsk_lock/details | base64 -d",
+      "stat mrsk_lock-app > /dev/null && cat mrsk_lock-app/details | base64 -d",
       new_command.status.join(" ")
   end
 
   test "acquire" do
     assert_match \
-      /mkdir mrsk_lock && echo ".*" > mrsk_lock\/details/m,
+      /mkdir mrsk_lock-app && echo ".*" > mrsk_lock-app\/details/m,
       new_command.acquire("Hello", "123").join(" ")
   end
 
   test "release" do
     assert_match \
-      "rm mrsk_lock/details && rm -r mrsk_lock",
+      "rm mrsk_lock-app/details && rm -r mrsk_lock-app",
       new_command.release.join(" ")
   end
 

test/commands/prune_test.rb

@@ -8,15 +8,21 @@ class CommandsPruneTest < ActiveSupport::TestCase
     }
   end
 
-  test "images" do
+  test "dangling images" do
     assert_equal \
-      "docker image prune --all --force --filter label=service=app --filter until=168h",
+      "docker image prune --force --filter label=service=app --filter dangling=true",
-      new_command.images.join(" ")
+      new_command.dangling_images.join(" ")
+  end
+
+  test "tagged images" do
+    assert_equal \
+      "docker image ls --filter label=service=app --format '{{.ID}} {{.Repository}}:{{.Tag}}' | grep -v -w \"$(docker container ls -a --format '{{.Image}}\\|' --filter label=service=app | tr -d '\\n')dhh/app:latest\\|dhh/app:<none>\" | while read image tag; do docker rmi $tag; done",
+      new_command.tagged_images.join(" ")
   end
 
   test "containers" do
     assert_equal \
-      "docker container prune --force --filter label=service=app --filter until=72h",
+      "docker ps -q -a --filter label=service=app --filter status=created --filter status=exited --filter status=dead | tail -n +6 | while read container_id; do docker rm $container_id; done",
       new_command.containers.join(" ")
   end
 

test/commands/traefik_test.rb

@@ -8,60 +8,77 @@ class CommandsTraefikTest < ActiveSupport::TestCase
       service: "app", image: "dhh/app", registry: { "username" => "dhh", "password" => "secret" }, servers: [ "1.1.1.1" ],
       traefik: { "image" => @image, "args" => { "accesslog.format" => "json", "api.insecure" => true, "metrics.prometheus.buckets" => "0.1,0.3,1.2,5.0" } }
     }
+
+    ENV["EXAMPLE_API_KEY"] = "456"
+  end
+
+  teardown do
+    ENV.delete("EXAMPLE_API_KEY")
   end
 
   test "run" do
     assert_equal \
-      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" #{@image} --providers.docker --log.level=DEBUG --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
+      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" #{@image} --providers.docker --log.level=\"DEBUG\" --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
       new_command.run.join(" ")
 
     @config[:traefik]["host_port"] = "8080"
     assert_equal \
-      "docker run --name traefik --detach --restart unless-stopped --publish 8080:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" #{@image} --providers.docker --log.level=DEBUG --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
+      "docker run --name traefik --detach --restart unless-stopped --publish 8080:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" #{@image} --providers.docker --log.level=\"DEBUG\" --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
       new_command.run.join(" ")
   end
 
   test "run with ports configured" do
     assert_equal \
-      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" #{@image} --providers.docker --log.level=DEBUG --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
+      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" #{@image} --providers.docker --log.level=\"DEBUG\" --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
       new_command.run.join(" ")
 
     @config[:traefik]["options"] = {"publish" => %w[9000:9000 9001:9001]}
     assert_equal \
-      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" --publish \"9000:9000\" --publish \"9001:9001\" #{@image} --providers.docker --log.level=DEBUG --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
+      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" --publish \"9000:9000\" --publish \"9001:9001\" #{@image} --providers.docker --log.level=\"DEBUG\" --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
       new_command.run.join(" ")
   end
 
   test "run with volumes configured" do
     assert_equal \
-      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" #{@image} --providers.docker --log.level=DEBUG --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
+      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" #{@image} --providers.docker --log.level=\"DEBUG\" --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
       new_command.run.join(" ")
 
     @config[:traefik]["options"] = {"volume" => %w[./letsencrypt/acme.json:/letsencrypt/acme.json] }
     assert_equal \
-      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" --volume \"./letsencrypt/acme.json:/letsencrypt/acme.json\" #{@image} --providers.docker --log.level=DEBUG --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
+      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" --volume \"./letsencrypt/acme.json:/letsencrypt/acme.json\" #{@image} --providers.docker --log.level=\"DEBUG\" --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
       new_command.run.join(" ")
   end
 
   test "run with several options configured" do
     assert_equal \
-      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" #{@image} --providers.docker --log.level=DEBUG --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
+      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" #{@image} --providers.docker --log.level=\"DEBUG\" --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
       new_command.run.join(" ")
 
     @config[:traefik]["options"] = {"volume" => %w[./letsencrypt/acme.json:/letsencrypt/acme.json], "publish" => %w[8080:8080], "memory" => "512m"}
     assert_equal \
-      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" --volume \"./letsencrypt/acme.json:/letsencrypt/acme.json\" --publish \"8080:8080\" --memory \"512m\" #{@image} --providers.docker --log.level=DEBUG --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
+      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" --volume \"./letsencrypt/acme.json:/letsencrypt/acme.json\" --publish \"8080:8080\" --memory \"512m\" #{@image} --providers.docker --log.level=\"DEBUG\" --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
       new_command.run.join(" ")
   end
 
   test "run with labels configured" do
     assert_equal \
-      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" #{@image} --providers.docker --log.level=DEBUG --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
+      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" #{@image} --providers.docker --log.level=\"DEBUG\" --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
       new_command.run.join(" ")
 
     @config[:traefik]["labels"] = { "traefik.http.routers.dashboard.service" => "api@internal", "traefik.http.routers.dashboard.middlewares" => "auth" }
     assert_equal \
-      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" --label traefik.http.routers.dashboard.service=\"api@internal\" --label traefik.http.routers.dashboard.middlewares=\"auth\" #{@image} --providers.docker --log.level=DEBUG --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
+      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" --label traefik.http.routers.dashboard.service=\"api@internal\" --label traefik.http.routers.dashboard.middlewares=\"auth\" #{@image} --providers.docker --log.level=\"DEBUG\" --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
+      new_command.run.join(" ")
+  end
+
+  test "run with env configured" do
+    assert_equal \
+      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" #{@image} --providers.docker --log.level=\"DEBUG\" --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
+      new_command.run.join(" ")
+
+    @config[:traefik]["env"] = { "secret" => %w[EXAMPLE_API_KEY] }
+    assert_equal \
+      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock -e EXAMPLE_API_KEY=\"456\" --log-opt max-size=\"10m\" #{@image} --providers.docker --log.level=\"DEBUG\" --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
       new_command.run.join(" ")
   end
 
@@ -69,7 +86,7 @@ class CommandsTraefikTest < ActiveSupport::TestCase
     @config.delete(:traefik)
 
     assert_equal \
-      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" #{Mrsk::Commands::Traefik::DEFAULT_IMAGE} --providers.docker --log.level=DEBUG",
+      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" #{Mrsk::Commands::Traefik::DEFAULT_IMAGE} --providers.docker --log.level=\"DEBUG\"",
       new_command.run.join(" ")
   end
 
@@ -77,7 +94,15 @@ class CommandsTraefikTest < ActiveSupport::TestCase
     @config[:logging] = { "driver" => "local", "options" => { "max-size" => "100m", "max-file" => "3" } }
 
     assert_equal \
-      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-driver \"local\" --log-opt max-size=\"100m\" --log-opt max-file=\"3\" #{@image} --providers.docker --log.level=DEBUG --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
+      "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-driver \"local\" --log-opt max-size=\"100m\" --log-opt max-file=\"3\" #{@image} --providers.docker --log.level=\"DEBUG\" --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
+      new_command.run.join(" ")
+  end
+
+  test "run with default args overriden" do
+    @config[:traefik]["args"]["log.level"] = "ERROR"
+
+    assert_equal \
+    "docker run --name traefik --detach --restart unless-stopped --publish 80:80 --volume /var/run/docker.sock:/var/run/docker.sock --log-opt max-size=\"10m\" #{@image} --providers.docker --log.level=\"ERROR\" --accesslog.format=\"json\" --api.insecure --metrics.prometheus.buckets=\"0.1,0.3,1.2,5.0\"",
       new_command.run.join(" ")
   end
 

test/configuration/builder_test.rb

@@ -0,0 +1,151 @@
+require "test_helper"
+
+class ConfigurationBuilderTest < ActiveSupport::TestCase
+  setup do
+    @deploy = {
+      service: "app", image: "dhh/app", registry: { "username" => "dhh", "password" => "secret" },
+      servers: [ "1.1.1.1" ]
+    }
+
+    @config = Mrsk::Configuration.new(@deploy)
+
+    @deploy_with_builder_option = {
+      service: "app", image: "dhh/app", registry: { "username" => "dhh", "password" => "secret" },
+      servers: [ "1.1.1.1" ],
+      builder: {}
+    }
+
+    @config_with_builder_option = Mrsk::Configuration.new(@deploy_with_builder_option)
+  end
+
+  test "multiarch?" do
+    assert_equal true, @config.builder.multiarch?
+  end
+
+  test "setting multiarch to false" do
+    @deploy_with_builder_option[:builder] = { "multiarch" => false }
+
+    assert_equal false, @config_with_builder_option.builder.multiarch?
+  end
+
+  test "local?" do
+    assert_equal false, @config.builder.local?
+  end
+
+  test "remote?" do
+    assert_equal false, @config.builder.remote?
+  end
+
+  test "remote_arch" do
+    assert_nil @config.builder.remote_arch
+  end
+
+  test "remote_host" do
+    assert_nil @config.builder.remote_host
+  end
+
+  test "remote config is missing when local is specified" do
+    @deploy_with_builder_option[:builder] = { "local" => { "arch" => "arm64", "host" => "unix:///Users/<%= `whoami`.strip %>/.docker/run/docker.sock" } }
+
+    assert_raises(ArgumentError) do
+      @config_with_builder_option.builder
+    end
+  end
+
+  test "setting both local and remote configs" do
+    @deploy_with_builder_option[:builder] = {
+      "local" => { "arch" => "arm64", "host" => "unix:///Users/<%= `whoami`.strip %>/.docker/run/docker.sock" },
+      "remote" => { "arch" => "amd64", "host" => "ssh://root@192.168.0.1" }
+    }
+
+    assert_equal true, @config_with_builder_option.builder.local?
+    assert_equal true, @config_with_builder_option.builder.remote?
+
+    assert_equal "amd64", @config_with_builder_option.builder.remote_arch
+    assert_equal "ssh://root@192.168.0.1", @config_with_builder_option.builder.remote_host
+
+    assert_equal "arm64", @config_with_builder_option.builder.local_arch
+    assert_equal "unix:///Users/<%= `whoami`.strip %>/.docker/run/docker.sock", @config_with_builder_option.builder.local_host
+  end
+
+  test "cached?" do
+    assert_equal false, @config.builder.cached?
+  end
+
+  test "invalid cache type specified" do
+    @deploy_with_builder_option[:builder] = { "cache" => { "type" => "invalid" } }
+
+    assert_raises(ArgumentError) do
+      @config_with_builder_option.builder
+    end
+  end
+
+  test "cache_from" do
+    assert_nil @config.builder.cache_from
+  end
+
+  test "cache_to" do
+    assert_nil @config.builder.cache_to
+  end
+
+  test "setting gha cache" do
+    @deploy_with_builder_option[:builder] = { "cache" => { "type" => "gha", "options" => "mode=max" } }
+
+    assert_equal "type=gha", @config_with_builder_option.builder.cache_from
+    assert_equal "type=gha,mode=max", @config_with_builder_option.builder.cache_to
+  end
+
+  test "setting registry cache" do
+    @deploy_with_builder_option[:builder] = { "cache" => { "type" => "registry", "options" => "mode=max,image-manifest=true,oci-mediatypes=true" } }
+
+    assert_equal "type=registry,ref=/dhh/app-build-cache", @config_with_builder_option.builder.cache_from
+    assert_equal "type=registry,mode=max,image-manifest=true,oci-mediatypes=true,ref=/dhh/app-build-cache", @config_with_builder_option.builder.cache_to
+  end
+
+  test "setting registry cache with image" do
+    @deploy_with_builder_option[:builder] = { "cache" => { "type" => "registry", "image" => "mrsk", "options" => "mode=max" } }
+
+    assert_equal "type=registry,ref=/mrsk", @config_with_builder_option.builder.cache_from
+    assert_equal "type=registry,mode=max,ref=/mrsk", @config_with_builder_option.builder.cache_to
+  end
+
+  test "args" do
+    assert_equal({}, @config.builder.args)
+  end
+
+  test "setting args" do
+    @deploy_with_builder_option[:builder] = { "args" => { "key" => "value" } }
+
+    assert_equal({ "key" => "value" }, @config_with_builder_option.builder.args)
+  end
+
+  test "secrets" do
+    assert_equal [], @config.builder.secrets
+  end
+
+  test "setting secrets" do
+    @deploy_with_builder_option[:builder] = { "secrets" => ["GITHUB_TOKEN"] }
+
+    assert_equal ["GITHUB_TOKEN"], @config_with_builder_option.builder.secrets
+  end
+
+  test "dockerfile" do
+    assert_equal "Dockerfile", @config.builder.dockerfile
+  end
+
+  test "setting dockerfile" do
+    @deploy_with_builder_option[:builder] = { "dockerfile" => "Dockerfile.dev" }
+
+    assert_equal "Dockerfile.dev", @config_with_builder_option.builder.dockerfile
+  end
+
+  test "context" do
+    assert_equal ".", @config.builder.context
+  end
+
+  test "setting context" do
+    @deploy_with_builder_option[:builder] = { "context" => ".." }
+
+    assert_equal "..", @config_with_builder_option.builder.context
+  end
+end

test/configuration/role_test.rb

@@ -42,7 +42,7 @@ class ConfigurationRoleTest < ActiveSupport::TestCase
   end
 
   test "special label args for web" do
-    assert_equal [ "--label", "service=\"app\"", "--label", "role=\"web\"", "--label", "traefik.http.routers.app-web.rule=\"PathPrefix(\\`/\\`)\"", "--label", "traefik.http.services.app-web.loadbalancer.healthcheck.path=\"/up\"", "--label", "traefik.http.services.app-web.loadbalancer.healthcheck.interval=\"1s\"", "--label", "traefik.http.middlewares.app-web-retry.retry.attempts=\"5\"", "--label", "traefik.http.middlewares.app-web-retry.retry.initialinterval=\"500ms\"", "--label", "traefik.http.routers.app-web.middlewares=\"app-web-retry@docker\"" ], @config.role(:web).label_args
+    assert_equal [ "--label", "service=\"app\"", "--label", "role=\"web\"", "--label", "traefik.http.services.app-web.loadbalancer.server.scheme=\"http\"", "--label", "traefik.http.routers.app-web.rule=\"PathPrefix(\\`/\\`)\"", "--label", "traefik.http.middlewares.app-web-retry.retry.attempts=\"5\"", "--label", "traefik.http.middlewares.app-web-retry.retry.initialinterval=\"500ms\"", "--label", "traefik.http.routers.app-web.middlewares=\"app-web-retry@docker\"" ], @config.role(:web).label_args
   end
 
   test "custom labels" do
@@ -57,8 +57,8 @@ class ConfigurationRoleTest < ActiveSupport::TestCase
   end
 
   test "overwriting default traefik label" do
-    @deploy[:labels] = { "traefik.http.routers.app.rule" => "\"Host(\\`example.com\\`) || (Host(\\`example.org\\`) && Path(\\`/traefik\\`))\"" }
+    @deploy[:labels] = { "traefik.http.routers.app-web.rule" => "\"Host(\\`example.com\\`) || (Host(\\`example.org\\`) && Path(\\`/traefik\\`))\"" }
-    assert_equal "\"Host(\\`example.com\\`) || (Host(\\`example.org\\`) && Path(\\`/traefik\\`))\"", @config.role(:web).labels["traefik.http.routers.app.rule"]
+    assert_equal "\"Host(\\`example.com\\`) || (Host(\\`example.org\\`) && Path(\\`/traefik\\`))\"", @config.role(:web).labels["traefik.http.routers.app-web.rule"]
   end
 
   test "default traefik label on non-web role" do
@@ -66,15 +66,7 @@ class ConfigurationRoleTest < ActiveSupport::TestCase
       c[:servers]["beta"] = { "traefik" => "true", "hosts" => [ "1.1.1.5" ] }
     })
 
-    assert_equal [ "--label", "service=\"app\"", "--label", "role=\"beta\"", "--label", "traefik.http.routers.app-beta.rule=\"PathPrefix(\\`/\\`)\"", "--label", "traefik.http.services.app-beta.loadbalancer.healthcheck.path=\"/up\"", "--label", "traefik.http.services.app-beta.loadbalancer.healthcheck.interval=\"1s\"", "--label", "traefik.http.middlewares.app-beta-retry.retry.attempts=\"5\"", "--label", "traefik.http.middlewares.app-beta-retry.retry.initialinterval=\"500ms\"", "--label", "traefik.http.routers.app-beta.middlewares=\"app-beta-retry@docker\"" ], config.role(:beta).label_args
+    assert_equal [ "--label", "service=\"app\"", "--label", "role=\"beta\"", "--label", "traefik.http.services.app-beta.loadbalancer.server.scheme=\"http\"", "--label", "traefik.http.routers.app-beta.rule=\"PathPrefix(\\`/\\`)\"", "--label", "traefik.http.middlewares.app-beta-retry.retry.attempts=\"5\"", "--label", "traefik.http.middlewares.app-beta-retry.retry.initialinterval=\"500ms\"", "--label", "traefik.http.routers.app-beta.middlewares=\"app-beta-retry@docker\"" ], config.role(:beta).label_args
-  end
-
-  test "default traefik label for non-web role with destination" do
-    config = Mrsk::Configuration.new(@deploy_with_roles.tap { |c|
-      c[:servers]["beta"] = { "traefik" => "true", "hosts" => [ "1.1.1.5" ] }
-    }, destination: "staging")
-
-    assert_equal [ "--label", "service=\"app\"", "--label", "role=\"beta\"", "--label", "destination=\"staging\"", "--label", "traefik.http.routers.app-beta-staging.rule=\"PathPrefix(\\`/\\`)\"", "--label", "traefik.http.services.app-beta-staging.loadbalancer.healthcheck.path=\"/up\"", "--label", "traefik.http.services.app-beta-staging.loadbalancer.healthcheck.interval=\"1s\"", "--label", "traefik.http.middlewares.app-beta-staging-retry.retry.attempts=\"5\"", "--label", "traefik.http.middlewares.app-beta-staging-retry.retry.initialinterval=\"500ms\"", "--label", "traefik.http.routers.app-beta-staging.middlewares=\"app-beta-staging-retry@docker\"" ], config.role(:beta).label_args
   end
 
   test "env overwritten by role" do

test/configuration_test.rb

@@ -48,7 +48,7 @@ class ConfigurationTest < ActiveSupport::TestCase
   end
 
   test "role" do
-    assert_equal "web", @config.role(:web).name
+    assert @config.role(:web).name.web?
     assert_equal "workers", @config_with_roles.role(:workers).name
     assert_nil @config.role(:missing)
   end
@@ -72,19 +72,36 @@ class ConfigurationTest < ActiveSupport::TestCase
     assert_equal [ "1.1.1.1", "1.1.1.2", "1.1.1.3" ], config.traefik_hosts
   end
 
-  test "version" do
+  test "version no git repo" do
     ENV.delete("VERSION")
 
     @config.expects(:system).with("git rev-parse").returns(nil)
     error = assert_raises(RuntimeError) { @config.version}
     assert_match /no git repository found/, error.message
+  end
 
-    @config.expects(:current_commit_hash).returns("git-version")
+  test "version from git committed" do
+    ENV.delete("VERSION")
+
+    @config.expects(:`).with("git rev-parse HEAD").returns("git-version")
+    @config.expects(:`).with("git status --porcelain").returns("")
     assert_equal "git-version", @config.version
+  end
 
+  test "version from git uncommitted" do
+    ENV.delete("VERSION")
+
+    @config.expects(:`).with("git rev-parse HEAD").returns("git-version")
+    @config.expects(:`).with("git status --porcelain").returns("M   file\n")
+    assert_match /^git-version_uncommitted_[0-9a-f]{16}$/, @config.version
+  end
+
+  test "version from env" do
     ENV["VERSION"] = "env-version"
     assert_equal "env-version", @config.version
+  end
 
+  test "version from arg" do
     @config.version = "arg-version"
     assert_equal "arg-version", @config.version
   end
@@ -149,7 +166,7 @@ class ConfigurationTest < ActiveSupport::TestCase
     assert_raises(KeyError) do
       config = Mrsk::Configuration.new(@deploy.tap { |c| c.merge!({
         env: { "secret" => [ "PASSWORD" ] }
-      }) })
+      }) }).ensure_env_available
     end
   end
 
@@ -194,17 +211,21 @@ class ConfigurationTest < ActiveSupport::TestCase
     assert_equal "root", @config.ssh_options[:user]
 
     config = Mrsk::Configuration.new(@deploy.tap { |c| c.merge!(ssh: { "user" => "app" }) })
-    assert_equal "app", @config.ssh_options[:user]
+    assert_equal "app", config.ssh_options[:user]
+    assert_equal 4, config.ssh_options[:logger].level
+
+    config = Mrsk::Configuration.new(@deploy.tap { |c| c.merge!(ssh: { "log_level" => "debug" }) })
+    assert_equal 0, config.ssh_options[:logger].level
   end
 
   test "ssh options with proxy host" do
     config = Mrsk::Configuration.new(@deploy.tap { |c| c.merge!(ssh: { "proxy" => "1.2.3.4" }) })
-    assert_equal "root@1.2.3.4", @config.ssh_options[:proxy].jump_proxies
+    assert_equal "root@1.2.3.4", config.ssh_options[:proxy].jump_proxies
   end
 
   test "ssh options with proxy host and user" do
     config = Mrsk::Configuration.new(@deploy.tap { |c| c.merge!(ssh: { "proxy" => "app@1.2.3.4" }) })
-    assert_equal "app@1.2.3.4", @config.ssh_options[:proxy].jump_proxies
+    assert_equal "app@1.2.3.4", config.ssh_options[:proxy].jump_proxies
   end
 
   test "volume_args" do
@@ -249,6 +270,37 @@ class ConfigurationTest < ActiveSupport::TestCase
   end
 
   test "to_h" do
-    assert_equal({ :roles=>["web"], :hosts=>["1.1.1.1", "1.1.1.2"], :primary_host=>"1.1.1.1", :version=>"missing", :repository=>"dhh/app", :absolute_image=>"dhh/app:missing", :service_with_version=>"app-missing", :env_args=>["-e", "REDIS_URL=\"redis://x/y\""], :ssh_options=>{:user=>"root", :auth_methods=>["publickey"]}, :volume_args=>["--volume", "/local/path:/container/path"], :logging=>["--log-opt", "max-size=\"10m\""], :healthcheck=>{"path"=>"/up", "port"=>3000, "max_attempts" => 7 }}, @config.to_h)
+    assert_equal({
+      :roles=>["web"],
+      :hosts=>["1.1.1.1", "1.1.1.2"],
+      :primary_host=>"1.1.1.1",
+      :version=>"missing",
+      :repository=>"dhh/app",
+      :absolute_image=>"dhh/app:missing",
+      :service_with_version=>"app-missing",
+      :env_args=>["-e", "REDIS_URL=\"redis://x/y\""],
+      :ssh_options=>{:user=>"root", :auth_methods=>["publickey"]},
+      :ssh_log_level=>4,
+      :volume_args=>["--volume", "/local/path:/container/path"],
+      :builder=>{},
+      :logging=>["--log-opt", "max-size=\"10m\""],
+      :healthcheck=>{"path"=>"/up", "port"=>3000, "max_attempts" => 7 }
+    }, @config.to_h)
+  end
+
+  test "min version is lower" do
+    config = Mrsk::Configuration.new(@deploy.tap { |c| c.merge!(minimum_version: "0.0.1") })
+    assert_equal "0.0.1", config.minimum_version
+  end
+
+  test "min version is equal" do
+    config = Mrsk::Configuration.new(@deploy.tap { |c| c.merge!(minimum_version: Mrsk::VERSION) })
+    assert_equal Mrsk::VERSION, config.minimum_version
+  end
+
+  test "min version is higher" do
+    assert_raises(ArgumentError) do
+      Mrsk::Configuration.new(@deploy.tap { |c| c.merge!(minimum_version: "10000.0.0") })
+    end
   end
 end

test/fixtures/deploy_with_boot_strategy.yml

@@ -0,0 +1,17 @@
+service: app
+image: dhh/app
+servers:
+  web:
+    - "1.1.1.1"
+    - "1.1.1.2"
+  workers:
+    - "1.1.1.3"
+    - "1.1.1.4"
+
+registry:
+  username: user
+  password: pw
+
+boot:
+  limit: 3
+  wait: 2

test/fixtures/deploy_with_percentage_boot_strategy.yml

@@ -0,0 +1,17 @@
+service: app
+image: dhh/app
+servers:
+  web:
+    - "1.1.1.1"
+    - "1.1.1.2"
+  workers:
+    - "1.1.1.3"
+    - "1.1.1.4"
+
+registry:
+  username: user
+  password: pw
+
+boot:
+  limit: 25%
+  wait: 2

test/fixtures/deploy_with_secrets.yml

@@ -0,0 +1,11 @@
+service: app
+image: dhh/app
+servers:
+  - "1.1.1.1"
+  - "1.1.1.2"
+registry:
+  username: user
+  password: pw
+env:
+  secret:
+    - PASSWORD

test/integration/accessory_test.rb

@@ -0,0 +1,36 @@
+require_relative "integration_test"
+
+class AccessoryTest < IntegrationTest
+  test "boot, stop, start, restart, logs, remove" do
+    mrsk :accessory, :boot, :busybox
+    assert_accessory_running :busybox
+
+    mrsk :accessory, :stop, :busybox
+    assert_accessory_not_running :busybox
+
+    mrsk :accessory, :start, :busybox
+    assert_accessory_running :busybox
+
+    mrsk :accessory, :restart, :busybox
+    assert_accessory_running :busybox
+
+    logs = mrsk :accessory, :logs, :busybox, capture: true
+    assert_match /Starting busybox.../, logs
+
+    mrsk :accessory, :remove, :busybox, "-y"
+    assert_accessory_not_running :busybox
+  end
+
+  private
+    def assert_accessory_running(name)
+      assert_match /registry:4443\/busybox:1.36.0   "sh -c 'echo \\"Start/, accessory_details(name)
+    end
+
+    def assert_accessory_not_running(name)
+      refute_match /registry:4443\/busybox:1.36.0   "sh -c 'echo \\"Start/, accessory_details(name)
+    end
+
+    def accessory_details(name)
+      mrsk :accessory, :details, name, capture: true
+    end
+end

test/integration/app_test.rb

@@ -0,0 +1,55 @@
+require_relative "integration_test"
+
+class AppTest < IntegrationTest
+  test "stop, start, boot, logs, images, containers, exec, remove" do
+    mrsk :deploy
+
+    assert_app_is_up
+
+    mrsk :app, :stop
+
+    # traefik is up and returns 404s when it can't match a route
+    assert_app_not_found
+
+    mrsk :app, :start
+
+    # mrsk app start does not wait
+    wait_for_app_to_be_up
+
+    mrsk :app, :boot
+
+    wait_for_app_to_be_up
+
+    logs = mrsk :app, :logs, capture: true
+    assert_match /App Host: vm1/, logs
+    assert_match /App Host: vm2/, logs
+    assert_match /GET \/ HTTP\/1.1/, logs
+
+    images = mrsk :app, :images, capture: true
+    assert_match /App Host: vm1/, images
+    assert_match /App Host: vm2/, images
+    assert_match /registry:4443\/app\s+#{latest_app_version}/, images
+    assert_match /registry:4443\/app\s+latest/, images
+
+    containers = mrsk :app, :containers, capture: true
+    assert_match /App Host: vm1/, containers
+    assert_match /App Host: vm2/, containers
+    assert_match /registry:4443\/app:#{latest_app_version}/, containers
+    assert_match /registry:4443\/app:latest/, containers
+
+    exec_output = mrsk :app, :exec, :ps, capture: true
+    assert_match /App Host: vm1/, exec_output
+    assert_match /App Host: vm2/, exec_output
+    assert_match /1 root      0:\d\d ps/, exec_output
+
+    exec_output = mrsk :app, :exec, "--reuse", :ps, capture: true
+    assert_match /App Host: vm1/, exec_output
+    assert_match /App Host: vm2/, exec_output
+    assert_match /1 root      0:\d\d nginx/, exec_output
+
+    mrsk :app, :remove
+
+    # traefik is up and returns 404s when it can't match a route
+    assert_app_not_found
+  end
+end

test/integration/docker-compose.yml

@@ -0,0 +1,57 @@
+version: "3.7"
+name: "mrsk-test"
+
+volumes:
+  shared:
+  registry:
+  deployer_bundle:
+
+services:
+  shared:
+    build:
+      context: docker/shared
+    volumes:
+      - shared:/shared
+
+  deployer:
+    privileged: true
+    build:
+      context: docker/deployer
+    environment:
+      - TEST_ID=${TEST_ID}
+    volumes:
+      - ../..:/mrsk
+      - shared:/shared
+      - registry:/registry
+      - deployer_bundle:/usr/local/bundle/
+
+  registry:
+    build:
+      context: docker/registry
+    environment:
+      - REGISTRY_HTTP_ADDR=0.0.0.0:4443
+      - REGISTRY_HTTP_TLS_CERTIFICATE=/certs/domain.crt
+      - REGISTRY_HTTP_TLS_KEY=/certs/domain.key
+    volumes:
+      - shared:/shared
+      - registry:/var/lib/registry/
+
+  vm1:
+    privileged: true
+    build:
+      context: docker/vm
+    volumes:
+      - shared:/shared
+
+  vm2:
+    privileged: true
+    build:
+      context: docker/vm
+    volumes:
+      - shared:/shared
+
+  load_balancer:
+    build:
+      context: docker/load_balancer
+    ports:
+      - "12345:80"

test/integration/docker/deployer/Dockerfile

@@ -0,0 +1,30 @@
+FROM ruby:3.2
+
+WORKDIR /app
+
+RUN apt-get update --fix-missing && apt-get install -y ca-certificates openssh-client curl gnupg docker.io
+
+RUN install -m 0755 -d /etc/apt/keyrings
+RUN curl -fsSL https://download.docker.com/linux/debian/gpg | gpg --dearmor -o /etc/apt/keyrings/docker.gpg
+RUN chmod a+r /etc/apt/keyrings/docker.gpg
+RUN echo \
+  "deb [arch="$(dpkg --print-architecture)" signed-by=/etc/apt/keyrings/docker.gpg] https://download.docker.com/linux/debian \
+  "$(. /etc/os-release && echo "$VERSION_CODENAME")" stable" | \
+  tee /etc/apt/sources.list.d/docker.list > /dev/null
+
+RUN apt-get update --fix-missing && apt-get install -y docker-ce docker-ce-cli containerd.io docker-buildx-plugin docker-compose-plugin
+
+COPY *.sh .
+COPY app/ .
+
+RUN rm -rf /root/.ssh
+RUN ln -s /shared/ssh /root/.ssh
+RUN mkdir -p /etc/docker/certs.d/registry:4443 && ln -s /shared/certs/domain.crt /etc/docker/certs.d/registry:4443/ca.crt
+
+RUN git config --global user.email "deployer@example.com"
+RUN git config --global user.name "Deployer"
+RUN git init && git add . && git commit -am "Initial version"
+
+HEALTHCHECK --interval=1s CMD pgrep sleep
+
+CMD ["./boot.sh"]

test/integration/docker/deployer/app/.env.erb

@@ -0,0 +1 @@
+SECRET_TOKEN=1234

test/integration/docker/deployer/app/.mrsk/hooks/post-deploy

@@ -0,0 +1,3 @@
+#!/bin/sh
+echo "Deployed!"
+mkdir -p /tmp/${TEST_ID} && touch /tmp/${TEST_ID}/post-deploy

test/integration/docker/deployer/app/.mrsk/hooks/pre-build

@@ -0,0 +1,3 @@
+#!/bin/sh
+echo "About to build and push..."
+mkdir -p /tmp/${TEST_ID} && touch /tmp/${TEST_ID}/pre-build

test/integration/docker/deployer/app/.mrsk/hooks/pre-connect

@@ -0,0 +1,8 @@
+#!/bin/sh
+
+echo "About to lock..."
+if [ "$MRSK_HOSTS" != "vm1,vm2" ]; then
+  echo "Expected hosts to be 'vm1,vm2', got $MRSK_HOSTS"
+  exit 1
+fi
+mkdir -p /tmp/${TEST_ID} && touch /tmp/${TEST_ID}/pre-connect

test/integration/docker/deployer/app/.mrsk/hooks/pre-deploy

@@ -0,0 +1,3 @@
+#!/bin/sh
+echo "Deployed!"
+mkdir -p /tmp/${TEST_ID} && touch /tmp/${TEST_ID}/pre-deploy

test/integration/docker/deployer/app/Dockerfile

@@ -0,0 +1,7 @@
+FROM registry:4443/nginx:1-alpine-slim
+
+COPY default.conf /etc/nginx/conf.d/default.conf
+
+ARG COMMIT_SHA
+RUN echo $COMMIT_SHA > /usr/share/nginx/html/version
+

test/integration/docker/deployer/app/config/deploy.yml

@@ -0,0 +1,27 @@
+service: app
+image: app
+servers:
+  - vm1
+  - vm2
+registry:
+  server: registry:4443
+  username: root
+  password: root
+builder:
+  multiarch: false
+  args:
+    COMMIT_SHA: <%= `git rev-parse HEAD` %>
+healthcheck:
+  cmd: wget -qO- http://localhost > /dev/null
+traefik:
+  args:
+    accesslog: true
+    accesslog.format: json
+  image: registry:4443/traefik:v2.9
+accessories:
+  busybox:
+    image: registry:4443/busybox:1.36.0
+    cmd: sh -c 'echo "Starting busybox..."; trap exit term; while true; do sleep 1; done'
+    roles:
+      - web
+stop_wait_time: 1

test/integration/docker/deployer/app/default.conf

@@ -0,0 +1,17 @@
+server {
+    listen       80;
+    listen  [::]:80;
+    server_name  localhost;
+
+    location / {
+        root   /usr/share/nginx/html;
+        index  index.html index.htm;
+    }
+
+    # redirect server error pages to the static page /50x.html
+    #
+    error_page   500 502 503 504  /50x.html;
+    location = /50x.html {
+        root   /usr/share/nginx/html;
+    }
+}

test/integration/docker/deployer/boot.sh

@@ -0,0 +1,5 @@
+#!/bin/bash
+
+dockerd --max-concurrent-downloads 1 &
+
+exec sleep infinity

test/integration/docker/deployer/setup.sh

@@ -0,0 +1,23 @@
+#!/bin/bash
+
+install_mrsk() {
+  cd /mrsk && gem build mrsk.gemspec -o /tmp/mrsk.gem && gem install /tmp/mrsk.gem
+}
+
+# Push the images to a persistent volume on the registry container
+# This is to work around docker hub rate limits
+push_image_to_registry_4443() {
+  # Check if the image is in the registry without having to pull it
+  if ! stat /registry/docker/registry/v2/repositories/$1/_manifests/tags/$2/current/link > /dev/null; then
+    hub_tag=$1:$2
+    registry_4443_tag=registry:4443/$1:$2
+    docker pull $hub_tag
+    docker tag $hub_tag $registry_4443_tag
+    docker push $registry_4443_tag
+  fi
+}
+
+install_mrsk
+push_image_to_registry_4443 nginx 1-alpine-slim
+push_image_to_registry_4443 traefik v2.9
+push_image_to_registry_4443 busybox 1.36.0

test/integration/docker/deployer/update_app_rev.sh

@@ -0,0 +1,3 @@
+#!/bin/bash
+
+git commit -am 'Update rev' --amend

test/integration/docker/load_balancer/Dockerfile

@@ -0,0 +1,5 @@
+FROM nginx:1-alpine-slim
+
+COPY default.conf /etc/nginx/conf.d/default.conf
+
+HEALTHCHECK --interval=1s CMD pgrep nginx

test/integration/docker/load_balancer/default.conf

@@ -0,0 +1,16 @@
+upstream loadbalancer {
+  server vm1:80;
+  server vm2:80;
+}
+
+server {
+  listen 80;
+
+  location / {
+    proxy_pass http://loadbalancer;
+    proxy_connect_timeout       10;
+    proxy_send_timeout          10;
+    proxy_read_timeout          10;
+    send_timeout                10;
+  }
+}

test/integration/docker/registry/Dockerfile

@@ -0,0 +1,9 @@
+FROM registry
+
+COPY boot.sh .
+
+RUN ln -s /shared/certs /certs
+
+HEALTHCHECK --interval=1s CMD pgrep registry
+
+ENTRYPOINT ["./boot.sh"]

test/integration/docker/registry/boot.sh

@@ -0,0 +1,5 @@
+#!/bin/sh
+
+while [ ! -f /certs/domain.crt ]; do sleep 1; done
+
+exec /entrypoint.sh /etc/docker/registry/config.yml

test/integration/docker/shared/Dockerfile

@@ -0,0 +1,17 @@
+FROM ubuntu:22.10
+
+WORKDIR /work
+
+RUN apt-get update --fix-missing && apt-get -y install openssh-client openssl
+
+RUN mkdir ssh && \
+  ssh-keygen -t rsa -f ssh/id_rsa -N ""
+
+COPY registry-dns.conf .
+COPY boot.sh .
+
+RUN mkdir certs && openssl req -newkey rsa:4096 -nodes -sha256 -keyout certs/domain.key   -x509 -days 365 -out certs/domain.crt   -subj '/CN=registry' -extensions EXT -config registry-dns.conf
+
+HEALTHCHECK --interval=1s CMD pgrep sleep
+
+CMD ["./boot.sh"]

test/integration/docker/shared/boot.sh

@@ -0,0 +1,5 @@
+#!/bin/bash
+
+cp -r * /shared
+
+exec sleep infinity

test/integration/docker/shared/registry-dns.conf

@@ -0,0 +1,7 @@
+[dn]
+CN=registry
+[req]
+distinguished_name = dn
+[EXT]
+subjectAltName=DNS:registry
+keyUsage=digitalSignature

test/integration/docker/vm/Dockerfile

@@ -0,0 +1,14 @@
+FROM ubuntu:22.10
+
+WORKDIR /work
+
+RUN apt-get update --fix-missing && apt-get -y install openssh-client openssh-server docker.io
+
+RUN mkdir /root/.ssh && ln -s /shared/ssh/id_rsa.pub /root/.ssh/authorized_keys
+RUN mkdir -p /etc/docker/certs.d/registry:4443 && ln -s /shared/certs/domain.crt /etc/docker/certs.d/registry:4443/ca.crt
+
+COPY boot.sh .
+
+HEALTHCHECK --interval=1s CMD pgrep dockerd
+
+CMD ["./boot.sh"]

test/integration/docker/vm/boot.sh

@@ -0,0 +1,9 @@
+#!/bin/bash
+
+while [ ! -f /root/.ssh/authorized_keys ]; do echo "Waiting for ssh keys"; sleep 1; done
+
+service ssh restart
+
+dockerd --max-concurrent-downloads 1 &
+
+exec sleep infinity

test/integration/integration_test.rb

@@ -0,0 +1,138 @@
+require "net/http"
+require "test_helper"
+
+class IntegrationTest < ActiveSupport::TestCase
+  setup do
+    ENV["TEST_ID"] = SecureRandom.hex
+    docker_compose "up --build -d"
+    wait_for_healthy
+    setup_deployer
+  end
+
+  teardown do
+    unless passed?
+      [:deployer, :vm1, :vm2, :shared, :load_balancer, :registry].each do |container|
+        puts
+        puts "Logs for #{container}:"
+        docker_compose :logs, container
+      end
+    end
+    docker_compose "down -t 1"
+  end
+
+  private
+    def docker_compose(*commands, capture: false, raise_on_error: true)
+      command = "TEST_ID=#{ENV["TEST_ID"]} docker compose #{commands.join(" ")}"
+      succeeded = false
+      if capture
+        result = stdouted { succeeded = system("cd test/integration && #{command}") }
+      else
+        succeeded = system("cd test/integration && #{command}")
+      end
+
+      raise "Command `#{command}` failed with error code `#{$?}`" if !succeeded && raise_on_error
+      result
+    end
+
+    def deployer_exec(*commands, **options)
+      docker_compose("exec deployer #{commands.join(" ")}", **options)
+    end
+
+    def mrsk(*commands, **options)
+      deployer_exec(:mrsk, *commands, **options)
+    end
+
+    def assert_app_is_down
+      response = app_response
+      debug_response_code(response, "502")
+      assert_equal "502", response.code
+    end
+
+    def assert_app_is_up(version: nil)
+      response = app_response
+      debug_response_code(response, "200")
+      assert_equal "200", response.code
+      assert_app_version(version, response) if version
+    end
+
+    def assert_app_not_found
+      response = app_response
+      debug_response_code(response, "404")
+      assert_equal "404", response.code
+    end
+
+    def wait_for_app_to_be_up(timeout: 20, up_count: 3)
+      timeout_at = Time.now + timeout
+      up_times = 0
+      response = app_response
+      while up_times < up_count && timeout_at > Time.now
+        sleep 0.1
+        up_times += 1 if response.code == "200"
+        response = app_response
+      end
+      assert_equal up_times, up_count
+    end
+
+    def app_response
+      Net::HTTP.get_response(URI.parse("http://localhost:12345/version"))
+    end
+
+    def update_app_rev
+      deployer_exec "./update_app_rev.sh"
+      latest_app_version
+    end
+
+    def latest_app_version
+      deployer_exec("git rev-parse HEAD", capture: true)
+    end
+
+    def assert_app_version(version, response)
+      assert_equal version, response.body.strip
+    end
+
+    def assert_hooks_ran(*hooks)
+      hooks.each do |hook|
+        file = "/tmp/#{ENV["TEST_ID"]}/#{hook}"
+        assert_equal "removed '#{file}'", deployer_exec("rm -v #{file}", capture: true).strip
+      end
+    end
+
+    def assert_200(response)
+      code = response.code
+      if code != "200"
+        puts "Got response code #{code}, here are the traefik logs:"
+        mrsk :traefik, :logs
+        puts "And here are the load balancer logs"
+        docker_compose :logs, :load_balancer
+        puts "Tried to get the response code again and got #{app_response.code}"
+      end
+      assert_equal "200", code
+    end
+
+    def wait_for_healthy(timeout: 20)
+      timeout_at = Time.now + timeout
+      while docker_compose("ps -a | tail -n +2 | grep -v '(healthy)' | wc -l", capture: true) != "0"
+        if timeout_at < Time.now
+          docker_compose("ps -a | tail -n +2 | grep -v '(healthy)'")
+          raise "Container not healthy after #{timeout} seconds" if timeout_at < Time.now
+        end
+        sleep 0.1
+      end
+    end
+
+    def setup_deployer
+      deployer_exec("./setup.sh") unless $DEPLOYER_SETUP
+      $DEPLOYER_SETUP = true
+    end
+
+    def debug_response_code(app_response, expected_code)
+      code = app_response.code
+      if code != expected_code
+        puts "Got response code #{code}, here are the traefik logs:"
+        mrsk :traefik, :logs
+        puts "And here are the load balancer logs"
+        docker_compose :logs, :load_balancer
+        puts "Tried to get the response code again and got #{app_response.code}"
+      end
+    end
+end

test/integration/lock_test.rb

@@ -0,0 +1,18 @@
+require_relative "integration_test"
+
+class LockTest < IntegrationTest
+  test "acquire, release, status" do
+    mrsk :lock, :acquire, "-m 'Integration Tests'"
+
+    status = mrsk :lock, :status, capture: true
+    assert_match /Locked by: Deployer at .*\nVersion: #{latest_app_version}\nMessage: Integration Tests/m, status
+
+    error = mrsk :deploy, capture: true, raise_on_error: false
+    assert_match /Deploy lock found/m, error
+
+    mrsk :lock, :release
+
+    status = mrsk :lock, :status, capture: true
+    assert_match /There is no deploy lock/m, status
+  end
+end

test/integration/main_test.rb

@@ -0,0 +1,59 @@
+require_relative "integration_test"
+
+class MainTest < IntegrationTest
+  test "deploy, redeploy, rollback, details and audit" do
+    first_version = latest_app_version
+
+    assert_app_is_down
+
+    mrsk :deploy
+    assert_app_is_up version: first_version
+    assert_hooks_ran "pre-connect", "pre-build", "pre-deploy", "post-deploy"
+
+    second_version = update_app_rev
+
+    mrsk :redeploy
+    assert_app_is_up version: second_version
+    assert_hooks_ran "pre-connect", "pre-build", "pre-deploy", "post-deploy"
+
+    mrsk :rollback, first_version
+    assert_hooks_ran "pre-connect", "pre-deploy", "post-deploy"
+    assert_app_is_up version: first_version
+
+    details = mrsk :details, capture: true
+    assert_match /Traefik Host: vm1/, details
+    assert_match /Traefik Host: vm2/, details
+    assert_match /App Host: vm1/, details
+    assert_match /App Host: vm2/, details
+    assert_match /traefik:v2.9/, details
+    assert_match /registry:4443\/app:#{first_version}/, details
+
+    audit = mrsk :audit, capture: true
+    assert_match /Booted app version #{first_version}.*Booted app version #{second_version}.*Booted app version #{first_version}.*/m, audit
+  end
+
+  test "envify" do
+    mrsk :envify
+
+    assert_equal "SECRET_TOKEN=1234", deployer_exec("cat .env", capture: true)
+  end
+
+  test "config" do
+    config = YAML.load(mrsk(:config, capture: true))
+    version = latest_app_version
+
+    assert_equal [ "web" ], config[:roles]
+    assert_equal [ "vm1", "vm2" ], config[:hosts]
+    assert_equal "vm1", config[:primary_host]
+    assert_equal version, config[:version]
+    assert_equal "registry:4443/app", config[:repository]
+    assert_equal "registry:4443/app:#{version}", config[:absolute_image]
+    assert_equal "app-#{version}", config[:service_with_version]
+    assert_equal [], config[:env_args]
+    assert_equal [], config[:volume_args]
+    assert_equal({ user: "root", auth_methods: [ "publickey" ] }, config[:ssh_options])
+    assert_equal({ "multiarch" => false, "args" => { "COMMIT_SHA" => version } }, config[:builder])
+    assert_equal [ "--log-opt", "max-size=\"10m\"" ], config[:logging]
+    assert_equal({ "path" => "/up", "port" => 3000, "max_attempts" => 7, "cmd" => "wget -qO- http://localhost > /dev/null" }, config[:healthcheck])
+  end
+end

test/integration/traefik_test.rb

@@ -0,0 +1,36 @@
+require_relative "integration_test"
+
+class TraefikTest < IntegrationTest
+  test "boot, stop, start, restart, logs, remove" do
+    mrsk :traefik, :boot
+    assert_traefik_running
+
+    mrsk :traefik, :stop
+    assert_traefik_not_running
+
+    mrsk :traefik, :start
+    assert_traefik_running
+
+    mrsk :traefik, :restart
+    assert_traefik_running
+
+    logs = mrsk :traefik, :logs, capture: true
+    assert_match /Traefik version [\d.]+ built on/, logs
+
+    mrsk :traefik, :remove
+    assert_traefik_not_running
+  end
+
+  private
+    def assert_traefik_running
+      assert_match /traefik:v2.9   "\/entrypoint.sh/, traefik_details
+    end
+
+    def assert_traefik_not_running
+      refute_match /traefik:v2.9   "\/entrypoint.sh/, traefik_details
+    end
+
+    def traefik_details
+      mrsk :traefik, :details, capture: true
+    end
+end