leifcr / auto-deploy-rails

gitlab auto-deploy chart to use with auto devops pipeline

Auto-deploy rails chart

Initially forked from GitLab's Auto-deploy Helm Chart (https://gitlab.com/gitlab-org/charts/auto-deploy-app), but due to need workers, this fork was created.

Initially sidekiq is implemented.

Delayed job should be a matter of adding the proper commands

Note: Worker has been added initially to gitlabs own chart: See this commit (https://gitlab.com/gitlab-org/charts/auto-deploy-app/commit/a76e282d6cb72b8f52315efc3fd933beb5aa2789)

v0.3.x release

v0.3.10 introduces AntiAffinity for workers and rails main service to limit 1 pod on each node. If you dont want this behaviour, you can disable it in setting podAntiAffinity or worker.podAntiAffinity to false

v0.3.x changes to use apiVersions compatible with kubernetes >= 0.14. All kubernetes clusters with versions below that will not work with v0.3.x

When upgrading from v0.2.x to v0.3.x chart, you might need to purge your deployment if it fails upgrading. Always deploy to testing/staging before production.

Requirements

• Helm 2.9.0 and above is required in order support "helm.sh/hook-delete-policy": before-hook-creation for migrations

securityContext (Important)

This is set to run as user 1001, and group 1001, so ensure that your container has user 1001 and 1001 created. See https://github.com/leifcr/rails5-kubernetes for example

Usage with helm

With gitlab

In your CI env variables, or in your modified .gitlab-ci.yml set the following:

Set AUTO_DEVOPS_CHART_REPOSITORY to https://leifcr.gitlab.io/auto-deploy-rails Set AUTO_DEVOPS_CHART to gitlab/auto-deploy-rails

Note: due to a bug in the .gitlab-ci.yml, you need to use gitlab/auto-deploy-rails instead of auto-deploy-rails/auto-deploy-rails, even if the latter makes sense looking at the repo. This will conflict if you are using gitlab charts on the same helm instance. As you are unlikely to use gitlab charts during deployment, it should work. If you use gitlab charts as well, look at the solution below.

You can also opt to change download_chart in gitlab-ci.yml to the following:

function download_chart() {
  if [[ ! -d chart ]]; then
    auto_chart=${AUTO_DEVOPS_CHART:-gitlab/auto-deploy-app}
    auto_chart_base=$(dirname $auto_chart)
    auto_chart_name=$(basename $auto_chart)
    auto_chart_name=${auto_chart_name%.tgz}
    auto_chart_name=${auto_chart_name%.tar.gz}
  else
    auto_chart="chart"
    auto_chart_name="chart"
  fi

  helm init --client-only
  helm repo add $auto_chart_base ${AUTO_DEVOPS_CHART_REPOSITORY:-https://charts.gitlab.io}
  if [[ ! -d "$auto_chart" ]]; then
    helm fetch ${auto_chart} --untar
  fi
  if [ "$auto_chart_name" != "chart" ]; then
    mv ${auto_chart_name} chart
  fi

  helm dependency update chart/
  helm dependency build chart/
}

Manual Usage

Add the repo

helm repo add auto-deploy-rails https://leifcr.gitlab.io/auto-deploy-rails/

Copy values.yaml from the repository and modify to fit your app.

Configuration

Parameter	Description	Default
service.replicaCount		1
service.revisionHistoryLimit		10
service.securityContext.enabled		true
service.securityContext.fsGroup		1001
service.securityContext.runAsUser		1001
image.repository		gitlab.example.com/group/project
image.tag		stable
image.pullPolicy		Always
image.secrets		[name: gitlab-registry]
podAnnotations	Pod annotations	{}
podAntiAffinity	If set to false, anti-affinity is disabled, allowing multiple rails pods to run on same node	true
application.track		stable
application.tier		web
application.migrateCommand	If present, this variable will run as a shell command within an application Container as a Helm pre-upgrade Hook. Intended to run migration commands.	nil
application.initializeCommand	If present, this variable will run as shall command within an application Container as a Helm post-install Hook. Intended to run database initialization commands.	nil
application.secretName	Pass in the name of a Secret which the deployment will load all key-value pairs from the Secret as environment variables in the application container.	nil
application.secretChecksum	Pass in the checksum of the secrets referenced by application.secretName.	nil
hpa.enabled	If true, enables horizontal pod autoscaler. A resource request is also required to be set, such as resources.requests.cpu: 200m.	false
hpa.minReplicas		1
hpa.maxReplicas		5
gitlab.env	GitLab environment.	nil
gitlab.envName	GitLab environment name.	nil
gitlab.envURL	GitLab environment URL.	nil
gitlab.app	GitLab project slug.	nil
service.enabled		true
service.args	Change the service arguments on the fly if needed	``
service.annotations	Service annotations	{}
service.name		web
service.type		ClusterIP
service.url		http://my.host.com/
service.additionalHosts	If present, this list will add additional hostnames to the server configuration.	nil
Alternative Names (SANs)	nil
service.externalPort		3000
service.internalPort		3000
ingress.tls.enabled	If true, enables SSL	true
ingress.tls.secretName	Name of the secret used to terminate SSL traffic	""
ingress.annotations	Ingress annotations	{}
livenessProbe.path	Path to access on the HTTP server on periodic probe of container liveness.	/
livenessProbe.scheme	Scheme to access the HTTP server (HTTP or HTTPS).	HTTP
livenessProbe.initialDelaySeconds	# of seconds after the container has started before liveness probes are initiated.	15
livenessProbe.timeoutSeconds	# of seconds after which the liveness probe times out.	15
readinessProbe.path	Path to access on the HTTP server on periodic probe of container readiness.	/
readinessProbe.scheme	Scheme to access the HTTP server (HTTP or HTTPS).	HTTP
readinessProbe.initialDelaySeconds	# of seconds after the container has started before readiness probes are initiated.	5
readinessProbe.timeoutSeconds	# of seconds after which the readiness probe times out.	3
worker.enabled		true
worker.podAntiAffinity	If set to false, affinity is disabled for workers, allowing multiple rails workers on each node	true
worker.securityContext.enabled		true
worker.securityContext.fsGroup		1001
worker.securityContext.runAsUser		1001
worker.args		bundle exec sidekiq
worker.replicaCount		1
worker.revisionHistoryLimit		10
worker.sidekiq_alive.enabled		true
worker.sidekiq_alive.livenessProbe.path	Path to access on the HTTP server on periodic probe of container liveness.	/
worker.sidekiq_alive.livenessProbe.initialDelaySeconds	# of seconds after the container has started before liveness probes are initiated.	15
worker.sidekiq_alive.livenessProbe.timeoutSeconds	# of seconds after which the liveness probe times out.	15
worker.sidekiq_alive.livenessProbe.port	Port for sidekiq_alive	7433
worker.sidekiq_alive.readinessProbe.path	Path to access on the HTTP server on periodic probe of container readiness.	/
worker.sidekiq_alive.readinessProbe.initialDelaySeconds	# of seconds after the container has started before readiness probes are initiated.	5
worker.sidekiq_alive.readinessProbe.port	Port for sidekiq_alive	7433
worker.sidekiq_alive.readinessProbe.timeoutSeconds	# of seconds after which the readiness probe times out.	3
postgresql.enabled		false
podDisruptionBudget.enabled		false
podDisruptionBudget.maxUnavailable		1
podDisruptionBudget.minAvailable	If present, this variable will configure minAvailable in the PodDisruptionBudget. ⚠️ if you have replicaCount: 1 and podDisruptionBudget.minAvailable: 1 kubectl drain will be blocked.	nil

NOTE: when providing ingress annotations, these must be givens as multiple --set commands OR set in a file

Example via set:

helm install . --dry-run --debug --set ingress.annotations."traefik\.ingress\.kubernetes\.io/redirect-entry-point"=\"https\" --set ingress.annotations."traefik\.ingress\.kubernetes\.io/redirect-permanent"=\"true\"

Optional, set it through a settings file and use -f. (See ingress_example_settings.yaml for example) helm install . --dry-run --debug -f ingress_example_settings.yaml

Static site generated with Jekyll. See chart_site folder for details