Remove the healthcheck step

To speed up deployments, we'll remove the healthcheck step. This adds some risk to deployments for non-web roles - if they don't have a Docker healthcheck configured then the only check we do is if the container is running. If there is a bad image we might see the container running before it exits and deploy it. Previously the healthcheck step would have avoided this by ensuring a web container could boot and serve traffic first. To mitigate this, we'll add a deployment barrier. Until one of the primary role containers passes its healthcheck, we'll keep the barrier up and avoid stopping the containers on the non-primary roles. It the primary role container fails its healthcheck, we'll close the barrier and shut down the new containers on the waiting roles. We also have a new integration test to check we correctly handle a a broken image. This highlighted that SSHKit's default runner will stop at the first error it encounters. We'll now have a custom runner that waits for all threads to finish allowing them to clean up.
2024-03-21 11:36:21 +00:00
parent 990f1b4413
commit 0efb5ccfff
24 changed files with 269 additions and 327 deletions
--- a/test/cli/app_test.rb
+++ b/test/cli/app_test.rb
@@ -118,6 +118,62 @@ class CliAppTest < CliTestCase
    end
  end

+  test "boot with web barrier opened" do
+    Object.any_instance.stubs(:sleep)
+
+    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$", "--quiet", "|", :xargs, :docker, :inspect, "--format", "'{{if .State.Health}}{{.State.Health.Status}}{{else}}{{.State.Status}}{{end}}'")
+      .returns("running").at_least_once # web health check passing
+
+    SSHKit::Backend::Abstract.any_instance.expects(:capture_with_info)
+      .with(:docker, :container, :ls, "--all", "--filter", "name=^app-web-123$", "--quiet", "|", :xargs, :docker, :inspect, "--format", "'{{if .State.Health}}{{.State.Health.Status}}{{else}}{{.State.Status}}{{end}}'")
+      .returns("unhealthy").at_least_once # web health check failing
+
+    SSHKit::Backend::Abstract.any_instance.expects(:capture_with_info)
+      .with(:docker, :container, :ls, "--all", "--filter", "name=^app-workers-latest$", "--quiet", "|", :xargs, :docker, :inspect, "--format", "'{{if .State.Health}}{{.State.Health.Status}}{{else}}{{.State.Status}}{{end}}'")
+      .returns("running").at_least_once # workers health check
+
+    run_command("boot", config: :with_roles, host: nil).tap do |output|
+      assert_match "Waiting at web barrier (1.1.1.3)...", output
+      assert_match "Waiting at web barrier (1.1.1.4)...", output
+      assert_match "Barrier opened (1.1.1.3)", output
+      assert_match "Barrier opened (1.1.1.4)", output
+    end
+  end
+
+  test "boot with web barrier closed" do
+    Thread.report_on_exception = false
+
+    Object.any_instance.stubs(:sleep)
+
+    SSHKit::Backend::Abstract.any_instance.stubs(:capture_with_info).returns("123") # old version
+
+    SSHKit::Backend::Abstract.any_instance.expects(:capture_with_info)
+      .with(:docker, :container, :ls, "--all", "--filter", "name=^app-web-latest$", "--quiet", "|", :xargs, :docker, :inspect, "--format", "'{{if .State.Health}}{{.State.Health.Status}}{{else}}{{.State.Status}}{{end}}'")
+      .returns("unhealthy").at_least_once # web health check failing
+
+    SSHKit::Backend::Abstract.any_instance.expects(:capture_with_info)
+      .with(:docker, :container, :ls, "--all", "--filter", "name=^app-workers-latest$", "--quiet", "|", :xargs, :docker, :inspect, "--format", "'{{if .State.Health}}{{.State.Health.Status}}{{else}}{{.State.Status}}{{end}}'")
+      .returns("running").at_least_once # workers health check passing
+
+    stderred do
+      run_command("boot", config: :with_roles, host: nil, allow_execute_error: true).tap do |output|
+        assert_match "Waiting at web barrier (1.1.1.3)...", output
+        assert_match "Waiting at web barrier (1.1.1.4)...", output
+        assert_match "Barrier closed, shutting down new container... (1.1.1.3)", output
+        assert_match "Barrier closed, shutting down new container... (1.1.1.4)", output
+        assert_match "Running docker container ls --all --filter name=^app-web-latest$ --quiet | xargs docker stop on 1.1.1.1", output
+        assert_match "Running docker container ls --all --filter name=^app-web-latest$ --quiet | xargs docker stop on 1.1.1.2", output
+        assert_match "Running docker container ls --all --filter name=^app-workers-latest$ --quiet | xargs docker stop on 1.1.1.3", output
+        assert_match "Running docker container ls --all --filter name=^app-workers-latest$ --quiet | xargs docker stop on 1.1.1.4", output
+      end
+    end
+  ensure
+    Thread.report_on_exception = true
+  end
+
  test "start" do
    run_command("start").tap do |output|
      assert_match "docker start app-web-999", output
@@ -283,8 +339,12 @@ class CliAppTest < CliTestCase
  end

  private
-    def run_command(*command, config: :with_accessories)
-      stdouted { Kamal::Cli::App.start([ *command, "-c", "test/fixtures/deploy_#{config}.yml", "--hosts", "1.1.1.1" ]) }
+    def run_command(*command, config: :with_accessories, host: "1.1.1.1", allow_execute_error: false)
+      stdouted do
+        Kamal::Cli::App.start([ *command, "-c", "test/fixtures/deploy_#{config}.yml", *([ "--hosts", host ] if host) ])
+      rescue SSHKit::Runner::ExecuteError => e
+        raise e unless allow_execute_error
+      end
    end

    def stub_running
--- a/test/cli/healthcheck_test.rb
+++ b/test/cli/healthcheck_test.rb
@@ -1,82 +0,0 @@
-require_relative "cli_test_case"
-
-class CliHealthcheckTest < CliTestCase
-  test "perform" do
-    # Prevent expected failures from outputting to terminal
-    Thread.report_on_exception = false
-
-    Kamal::Cli::Healthcheck::Poller.stubs(:sleep) # No sleeping when retrying
-    Kamal::Configuration.any_instance.stubs(:run_id).returns("12345678901234567890123456789012")
-
-    SSHKit::Backend::Abstract.any_instance.stubs(:execute)
-      .with(:docker, :container, :ls, "--all", "--filter", "name=^healthcheck-app-999$", "--quiet", "|", :xargs, :docker, :stop, raise_on_non_zero_exit: false)
-    SSHKit::Backend::Abstract.any_instance.stubs(:execute)
-      .with(:docker, :run, "--detach", "--name", "healthcheck-app-999", "--publish", "3999:3000", "--label", "service=healthcheck-app", "-e", "KAMAL_CONTAINER_NAME=\"healthcheck-app\"", "--env-file", ".kamal/env/roles/app-web.env", "--health-cmd", "\"curl -f http://localhost:3000/up || exit 1\"", "--health-interval", "\"1s\"", "dhh/app:999")
-    SSHKit::Backend::Abstract.any_instance.stubs(:execute)
-      .with(:docker, :container, :ls, "--all", "--filter", "name=^healthcheck-app-999$", "--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(:docker, :container, :ls, "--all", "--filter", "name=^healthcheck-app-999$", "--quiet", "|", :xargs, :docker, :inspect, "--format", "'{{if .State.Health}}{{.State.Health.Status}}{{else}}{{.State.Status}}{{end}}'")
-      .returns("starting")
-      .then
-      .returns("unhealthy")
-      .then
-      .returns("healthy")
-
-    run_command("perform").tap do |output|
-      assert_match "container not ready (starting), retrying in 1s (attempt 1/7)...", output
-      assert_match "container not ready (unhealthy), retrying in 2s (attempt 2/7)...", output
-      assert_match "Container is healthy!", output
-    end
-  end
-
-  test "perform failing to become healthy" do
-    # Prevent expected failures from outputting to terminal
-    Thread.report_on_exception = false
-
-    Kamal::Cli::Healthcheck::Poller.stubs(:sleep) # No sleeping when retrying
-
-    SSHKit::Backend::Abstract.any_instance.stubs(:execute)
-      .with(:docker, :container, :ls, "--all", "--filter", "name=^healthcheck-app-999$", "--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", "KAMAL_CONTAINER_NAME=\"healthcheck-app\"", "--env-file", ".kamal/env/roles/app-web.env", "--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$", "--quiet", "|", :xargs, :docker, :container, :rm, raise_on_non_zero_exit: false)
-
-    # Continually report unhealthy
-    SSHKit::Backend::Abstract.any_instance.stubs(:capture_with_info)
-      .with(:docker, :container, :ls, "--all", "--filter", "name=^healthcheck-app-999$", "--quiet", "|", :xargs, :docker, :inspect, "--format", "'{{if .State.Health}}{{.State.Health.Status}}{{else}}{{.State.Status}}{{end}}'")
-      .returns("unhealthy")
-
-    # Capture logs when failing
-    SSHKit::Backend::Abstract.any_instance.stubs(:capture_with_info)
-      .with(:docker, :container, :ls, "--all", "--filter", "name=^healthcheck-app-999$", "--quiet", "|", :xargs, :docker, :logs, "--tail", 50, "2>&1")
-      .returns("some log output")
-
-    # Capture container health log when failing
-    SSHKit::Backend::Abstract.any_instance.stubs(:capture_with_pretty_json)
-      .with(:docker, :container, :ls, "--all", "--filter", "name=^healthcheck-app-999$", "--quiet", "|", :xargs, :docker, :inspect, "--format",  "'{{json .State.Health}}'")
-      .returns('{"Status":"unhealthy","Log":[{"ExitCode": 1,"Output": "/bin/sh: 1: curl: not found\n"}]}"')
-
-    exception = assert_raises do
-      run_command("perform")
-    end
-    assert_match "container not ready (unhealthy)", exception.message
-  end
-
-  test "raises an exception if primary does not have traefik" do
-    SSHKit::Backend::Abstract.any_instance.expects(:execute).never
-
-    exception = assert_raises do
-      run_command("perform", config_file: "test/fixtures/deploy_workers_only.yml")
-    end
-
-    assert_equal "The primary host is not configured to run Traefik", exception.message
-  end
-
-  private
-    def run_command(*command, config_file: "test/fixtures/deploy_with_accessories.yml")
-      stdouted { Kamal::Cli::Healthcheck.start([ *command, "-c", config_file ]) }
-    end
-end
--- a/test/cli/main_test.rb
+++ b/test/cli/main_test.rb
@@ -27,7 +27,6 @@ class CliMainTest < CliTestCase
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:registry:login", [], invoke_options)
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:build:pull", [], invoke_options)
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:traefik:boot", [], invoke_options)
-    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:healthcheck:perform", [], invoke_options)
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:app:stale_containers", [], invoke_options.merge(stop: true))
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:app:boot", [], invoke_options)
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:prune:all", [], invoke_options)
@@ -40,7 +39,6 @@ class CliMainTest < CliTestCase
      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
@@ -53,7 +51,6 @@ class CliMainTest < CliTestCase
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:registry:login", [], invoke_options)
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:build:deliver", [], invoke_options)
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:traefik:boot", [], invoke_options)
-    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:healthcheck:perform", [], invoke_options)
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:app:stale_containers", [], invoke_options.merge(stop: true))
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:app:boot", [], invoke_options)
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:prune:all", [], invoke_options)
@@ -67,7 +64,6 @@ class CliMainTest < CliTestCase
      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: true
@@ -80,7 +76,6 @@ class CliMainTest < CliTestCase
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:registry:login", [], invoke_options)
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:build:pull", [], invoke_options)
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:traefik:boot", [], invoke_options)
-    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:healthcheck:perform", [], invoke_options)
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:app:stale_containers", [], invoke_options.merge(stop: true))
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:app:boot", [], invoke_options)
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:prune:all", [], invoke_options)
@@ -90,7 +85,6 @@ class CliMainTest < CliTestCase
      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
@@ -156,7 +150,6 @@ class CliMainTest < CliTestCase
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:registry:login", [], invoke_options)
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:build:deliver", [], invoke_options)
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:traefik:boot", [], invoke_options)
-    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:healthcheck:perform", [], invoke_options)
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:app:stale_containers", [], invoke_options.merge(stop: true))
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:app:boot", [], invoke_options)
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:prune:all", [], invoke_options)
@@ -187,7 +180,6 @@ class CliMainTest < CliTestCase
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:registry:login", [], invoke_options)
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:build:deliver", [], invoke_options)
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:traefik:boot", [], invoke_options)
-    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:healthcheck:perform", [], invoke_options)
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:app:stale_containers", [], invoke_options.merge(stop: true))
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:app:boot", [], invoke_options)
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:prune:all", [], invoke_options)
@@ -199,7 +191,6 @@ class CliMainTest < CliTestCase
    invoke_options = { "config_file" => "test/fixtures/deploy_simple.yml", "version" => "999", "skip_hooks" => false, "verbose" => true }

    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:build:deliver", [], invoke_options)
-    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:healthcheck:perform", [], invoke_options)
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:app:stale_containers", [], invoke_options.merge(stop: true))
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:app:boot", [], invoke_options)

@@ -212,7 +203,6 @@ class CliMainTest < CliTestCase
      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: true
    end
  end
@@ -221,13 +211,11 @@ class CliMainTest < CliTestCase
    invoke_options = { "config_file" => "test/fixtures/deploy_simple.yml", "version" => "999", "skip_hooks" => false }

    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:build:pull", [], invoke_options)
-    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:healthcheck:perform", [], invoke_options)
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:app:stale_containers", [], invoke_options.merge(stop: true))
    Kamal::Cli::Main.any_instance.expects(:invoke).with("kamal:cli:app:boot", [], invoke_options)

    run_command("redeploy", "--skip_push").tap do |output|
      assert_match /Pull app image/, output
-      assert_match /Ensure app can pass healthcheck/, output
    end
  end

--- a/test/commander_test.rb
+++ b/test/commander_test.rb
@@ -99,6 +99,11 @@ class CommanderTest < ActiveSupport::TestCase
    assert_equal [ "workers" ], @kamal.roles_on("1.1.1.3").map(&:name)
  end

+  test "roles_on web comes first" do
+    configure_with(:deploy_with_two_roles_one_host)
+    assert_equal [ "web", "workers" ], @kamal.roles_on("1.1.1.1").map(&:name)
+  end
+
  test "default group strategy" do
    assert_empty @kamal.boot_strategy
  end
--- a/test/commands/healthcheck_test.rb
+++ b/test/commands/healthcheck_test.rb
@@ -1,124 +0,0 @@
-require "test_helper"
-
-class CommandsHealthcheckTest < ActiveSupport::TestCase
-  setup do
-    @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" } }
-    }
-  end
-
-  test "run" do
-    assert_equal \
-      "docker run --detach --name healthcheck-app-123 --publish 3999:3000 --label service=healthcheck-app -e KAMAL_CONTAINER_NAME=\"healthcheck-app\" --env-file .kamal/env/roles/app-web.env --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 port" do
-    @config[:healthcheck] = { "port" => 3001 }
-
-    assert_equal \
-      "docker run --detach --name healthcheck-app-123 --publish 3999:3001 --label service=healthcheck-app -e KAMAL_CONTAINER_NAME=\"healthcheck-app\" --env-file .kamal/env/roles/app-web.env --health-cmd \"curl -f http://localhost:3001/up || exit 1\" --health-interval \"1s\" dhh/app:123",
-      new_command.run.join(" ")
-  end
-
-  test "run with destination" do
-    @destination = "staging"
-
-    assert_equal \
-      "docker run --detach --name healthcheck-app-staging-123 --publish 3999:3000 --label service=healthcheck-app-staging -e KAMAL_CONTAINER_NAME=\"healthcheck-app-staging\" --env-file .kamal/env/roles/app-web-staging.env --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 KAMAL_CONTAINER_NAME=\"healthcheck-app\" --env-file .kamal/env/roles/app-web.env --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" } } }
-    @config[:healthcheck] = { "exposed_port" => 4999 }
-    assert_equal \
-      "docker run --detach --name healthcheck-app-123 --publish 4999:3000 --label service=healthcheck-app -e KAMAL_CONTAINER_NAME=\"healthcheck-app\" --env-file .kamal/env/roles/app-web.env --health-cmd \"curl -f http://localhost:3000/up || exit 1\" --health-interval \"1s\" --mount \"somewhere\" dhh/app:123",
-      new_command.run.join(" ")
-  end
-
-  test "run with tags" do
-    @config[:servers] = [ { "1.1.1.1" => "tag1" } ]
-    @config[:env] = {}
-    @config[:env]["tags"] = { "tag1" => { "ENV1" => "value1" } }
-
-    assert_equal \
-      "docker run --detach --name healthcheck-app-123 --publish 3999:3000 --label service=healthcheck-app -e KAMAL_CONTAINER_NAME=\"healthcheck-app\" --env-file .kamal/env/roles/app-web.env --env ENV1=\"value1\" --health-cmd \"curl -f http://localhost:3000/up || exit 1\" --health-interval \"1s\" dhh/app:123",
-      new_command.run.join(" ")
-  end
-
-  test "status" do
-    assert_equal \
-      "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.status.join(" ")
-  end
-
-  test "container_health_log" do
-    assert_equal \
-      "docker container ls --all --filter name=^healthcheck-app-123$ --quiet | xargs docker inspect --format '{{json .State.Health}}'",
-      new_command.container_health_log.join(" ")
-  end
-
-  test "stop" do
-    assert_equal \
-      "docker container ls --all --filter name=^healthcheck-app-123$ --quiet | xargs docker stop",
-      new_command.stop.join(" ")
-  end
-
-  test "stop with destination" do
-    @destination = "staging"
-
-    assert_equal \
-      "docker container ls --all --filter name=^healthcheck-app-staging-123$ --quiet | xargs docker stop",
-      new_command.stop.join(" ")
-  end
-
-  test "remove" do
-    assert_equal \
-      "docker container ls --all --filter name=^healthcheck-app-123$ --quiet | xargs docker container rm",
-      new_command.remove.join(" ")
-  end
-
-  test "remove with destination" do
-    @destination = "staging"
-
-    assert_equal \
-      "docker container ls --all --filter name=^healthcheck-app-staging-123$ --quiet | xargs docker container rm",
-      new_command.remove.join(" ")
-  end
-
-  test "logs" do
-    assert_equal \
-      "docker container ls --all --filter name=^healthcheck-app-123$ --quiet | xargs docker logs --tail 50 2>&1",
-      new_command.logs.join(" ")
-  end
-
-  test "logs with custom lines number" do
-    @config[:healthcheck] = { "log_lines" => 150 }
-    assert_equal \
-      "docker container ls --all --filter name=^healthcheck-app-123$ --quiet | xargs docker logs --tail 150 2>&1",
-      new_command.logs.join(" ")
-  end
-
-  test "logs with destination" do
-    @destination = "staging"
-
-    assert_equal \
-      "docker container ls --all --filter name=^healthcheck-app-staging-123$ --quiet | xargs docker logs --tail 50 2>&1",
-      new_command.logs.join(" ")
-  end
-
-  private
-    def new_command
-      Kamal::Commands::Healthcheck.new(Kamal::Configuration.new(@config, destination: @destination, version: "123"))
-    end
-end
--- a/test/configuration_test.rb
+++ b/test/configuration_test.rb
@@ -272,7 +272,7 @@ class ConfigurationTest < ActiveSupport::TestCase
        volume_args: [ "--volume", "/local/path:/container/path" ],
        builder: {},
        logging: [ "--log-opt", "max-size=\"10m\"" ],
-        healthcheck: { "path"=>"/up", "port"=>3000, "max_attempts" => 7, "exposed_port" => 3999, "cord" => "/tmp/kamal-cord", "log_lines" => 50 } }
+        healthcheck: { "path"=>"/up", "port"=>3000, "max_attempts" => 7, "cord" => "/tmp/kamal-cord", "log_lines" => 50 } }

    assert_equal expected_config, @config.to_h
  end
--- a/test/fixtures/deploy_with_two_roles_one_host.yml
+++ b/test/fixtures/deploy_with_two_roles_one_host.yml
@@ -0,0 +1,15 @@
+service: app
+image: dhh/app
+servers:
+  workers:
+    hosts:
+      - 1.1.1.1
+  web:
+    hosts:
+      - 1.1.1.1
+env:
+  REDIS_URL: redis://x/y
+registry:
+  server: registry.digitalocean.com
+  username: user
+  password: pw
--- a/test/integration/broken_deploy_test.rb
+++ b/test/integration/broken_deploy_test.rb
@@ -0,0 +1,24 @@
+require_relative "integration_test"
+
+class BrokenDeployTest < IntegrationTest
+  test "deploying a bad image" do
+    @app = "app_with_roles"
+
+    kamal :envify
+
+    first_version = latest_app_version
+
+    kamal :deploy
+
+    assert_app_is_up version: first_version
+    assert_container_running host: :vm3, name: "app-workers-#{first_version}"
+
+    second_version = break_app
+
+    kamal :deploy, raise_on_error: false
+
+    assert_app_is_up version: first_version
+    assert_container_running host: :vm3, name: "app-workers-#{first_version}"
+    assert_container_not_running host: :vm3, name: "app-workers-#{second_version}"
+  end
+end
--- a/test/integration/docker/deployer/app/config/deploy.yml
+++ b/test/integration/docker/deployer/app/config/deploy.yml
@@ -28,6 +28,7 @@ builder:
    COMMIT_SHA: <%= `git rev-parse HEAD` %>
 healthcheck:
  cmd: wget -qO- http://localhost > /dev/null || exit 1
+  max_attempts: 3
 traefik:
  args:
    accesslog: true
@@ -41,3 +42,4 @@ accessories:
    roles:
      - web
 stop_wait_time: 1
+readiness_delay: 0
--- a/test/integration/docker/deployer/app_with_roles/config/deploy.yml
+++ b/test/integration/docker/deployer/app_with_roles/config/deploy.yml
@@ -22,6 +22,7 @@ builder:
    COMMIT_SHA: <%= `git rev-parse HEAD` %>
 healthcheck:
  cmd: wget -qO- http://localhost > /dev/null || exit 1
+  max_attempts: 3
 traefik:
  args:
    accesslog: true
@@ -35,3 +36,4 @@ accessories:
    roles:
      - web
 stop_wait_time: 1
+readiness_delay: 0
--- a/test/integration/docker/deployer/break_app.sh
+++ b/test/integration/docker/deployer/break_app.sh
@@ -0,0 +1,3 @@
+#!/bin/bash
+
+cd $1 && echo "bad nginx config" > default.conf && git commit -am 'Broken'
--- a/test/integration/integration_test.rb
+++ b/test/integration/integration_test.rb
@@ -78,6 +78,11 @@ class IntegrationTest < ActiveSupport::TestCase
      latest_app_version
    end

+    def break_app
+      deployer_exec "./break_app.sh #{@app}", workdir: "/"
+      latest_app_version
+    end
+
    def latest_app_version
      deployer_exec("git rev-parse HEAD", capture: true)
    end
@@ -131,4 +136,16 @@ class IntegrationTest < ActiveSupport::TestCase
        puts "Tried to get the response code again and got #{app_response.code}"
      end
    end
+
+    def assert_container_running(host:, name:)
+      assert container_running?(host: host, name: name)
+    end
+
+    def assert_container_not_running(host:, name:)
+      assert_not container_running?(host: host, name: name)
+    end
+
+    def container_running?(host:, name:)
+      docker_compose("exec #{host} docker ps --filter=name=#{name} | tail -n+2", capture: true).tap { |x| p [ x, x.strip, x.strip.present? ] }.strip.present?
+    end
 end
--- a/test/integration/main_test.rb
+++ b/test/integration/main_test.rb
@@ -56,6 +56,12 @@ class MainTest < IntegrationTest
    assert_app_is_up version: version
    assert_hooks_ran "pre-connect", "pre-build", "pre-deploy", "post-deploy"
    assert_container_running host: :vm3, name: "app-workers-#{version}"
+
+    second_version = update_app_rev
+
+    kamal :redeploy
+    assert_app_is_up version: second_version
+    assert_container_running host: :vm3, name: "app-workers-#{second_version}"
  end

  test "config" do
@@ -73,7 +79,7 @@ class MainTest < IntegrationTest
    assert_equal({ user: "root", port: 22, keepalive: true, keepalive_interval: 30, log_level: :fatal }, 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, "exposed_port" => 3999, "cord"=>"/tmp/kamal-cord", "log_lines" => 50, "cmd"=>"wget -qO- http://localhost > /dev/null || exit 1" }, config[:healthcheck])
+    assert_equal({ "path" => "/up", "port" => 3000, "max_attempts" => 3, "cord"=>"/tmp/kamal-cord", "log_lines" => 50, "cmd"=>"wget -qO- http://localhost > /dev/null || exit 1" }, config[:healthcheck])
  end

  test "setup and remove" do
@@ -157,8 +163,4 @@ class MainTest < IntegrationTest
      assert vm1_image_ids.any?
      assert vm1_container_ids.any?
    end
-
-    def assert_container_running(host:, name:)
-      assert docker_compose("exec #{host} docker ps --filter=name=#{name} -q", capture: true).strip.present?
-    end
 end