Using review apps in the development of GitLab

Review apps are deployed using the start-review-app-pipeline job which triggers a child pipeline containing a series of jobs to perform the various tasks needed to deploy a review app.

start-review-app-pipeline job

For any of the following scenarios, the start-review-app-pipeline job would be automatically started (only when the merge request is approved):

  • for merge requests with CI configuration changes
  • for merge requests with frontend changes
  • for merge requests with changes to {,ee/,jh/}{app/controllers}/**/*
  • for merge requests with changes to {,ee/,jh/}{app/models}/**/*
  • for merge requests with changes to {,ee/,jh/}lib/{,ee/,jh/}gitlab/**/*
  • for merge requests with QA changes
  • for scheduled pipelines
  • the MR has the pipeline:run-review-app label set

E2E test runs on review apps

On every pipeline in the qa stage (which comes after the review stage), the review-qa-smoke and review-qa-blocking jobs are automatically started.

qa stage consists of following jobs:

  • review-qa-smoke: small and fast subset of tests to validate core functionality of GitLab.
  • review-qa-blocking: subset of tests that block the merge request. These tests are considered stable and are not allowed to fail.
  • review-qa-non-blocking: rest of the e2e tests that can be triggered manually.

review-qa-* jobs ensure that end-to-end tests for the changes in the merge request pass in a live environment. This shifts the identification of e2e failures from an environment on the path to production to the merge request to prevent breaking features on GitLab.com or costly GitLab.com deployment blockers. If needed, review-qa-* failures should be investigated with an SET (software engineer in test) counterpart to help determine the root cause of the error.

After the end-to-end test runs have finished, Allure reports are generated and published by the e2e-test-report job. A comment with links to the reports is added to the merge request.

Errors can be found in the gitlab-review-apps Sentry project and filterable by review app URL or commit SHA.

Bypass failed review app deployment to merge a broken master fix

Maintainers can elect to use the process for merging during broken master if a customer-critical merge request is blocked by pipelines failing due to review app deployment failures.

Performance Metrics

On every Review App child pipeline in the qa stage, the browser_performance job is automatically started: this job does basic browser performance testing using a Sitespeed.io Container.

Sample Data for review apps

Upon deployment of a review app, project data is created from the sample-gitlab-project template project. This aims to provide projects with prepopulated resources to facilitate manual and exploratory testing.

The sample projects will be created in the root user namespace and can be accessed from the personal projects list for that user.

How to

Redeploy review app from a clean slate

To reset review app and redeploy from a clean slate, do the following:

  1. Run review-stop job.
  2. Re-deploy by running or retrying review-deploy job.

Doing this will remove all existing data from a previously deployed review app.

Get access to the GCP review apps cluster

You need to open an access request (internal link) for the gcp-review-apps-dev GCP group and role.

This grants you the following permissions for:

Sign in to my review app

For GitLab Team Members only. If you want to sign in to the review app, review the GitLab handbook information for the shared 1Password account.

  • The default username is root.
  • The password can be found in the 1Password login item named GitLab EE Review App.

Enable a feature flag for my review app

  1. Open your review app and sign in as documented above.
  2. Create a personal access token.
  3. Enable the feature flag using the Feature flag API.

Find my review app slug

  1. Open the review-deploy job.
  2. Look for ** Deploying review-*.
  3. For instance for ** Deploying review-1234-abc-defg... **, your review app slug would be review-1234-abc-defg in this case.

Run a Rails console

  1. Make sure you have access to the cluster and the container.pods.exec permission first.
  2. Filter Workloads by your review app slug. For example, review-qa-raise-e-12chm0.
  3. Find and open the toolbox Deployment. For example, review-qa-raise-e-12chm0-toolbox.
  4. Select the Pod in the “Managed pods” section. For example, review-qa-raise-e-12chm0-toolbox-d5455cc8-2lsvz.
  5. Select the KUBECTL dropdown list, then Exec -> toolbox.
  6. Replace -c toolbox -- ls with -it -- gitlab-rails console from the default command or
    • Run kubectl exec --namespace review-qa-raise-e-12chm0 review-qa-raise-e-12chm0-toolbox-d5455cc8-2lsvz -it -- gitlab-rails console and
      • Replace review-qa-raise-e-12chm0-toolbox-d5455cc8-2lsvz with your Pod’s name.

Dig into a Pod’s logs

  1. Make sure you have access to the cluster and the container.pods.getLogs permission first.
  2. Filter Workloads by your review app slug. For example, review-qa-raise-e-12chm0.
  3. Find and open the migrations Deployment. For example, review-qa-raise-e-12chm0-migrations.1.
  4. Select the Pod in the “Managed pods” section. For example, review-qa-raise-e-12chm0-migrations.1-nqwtx.
  5. Select Container logs.

Alternatively, you could use the Logs Explorer which provides more utility to search logs. An example query for a pod name is as follows: 

resource.labels.pod_name:"review-qa-raise-e-12chm0-migrations"

How does it work?

CI/CD architecture diagram

graph TD A["build-qa-image, compile-production-assets<br/>(canonical default refs only)"]; B1[start-review-app-pipeline]; B[review-build-cng]; C["review-deploy<br><br>Helm deploys the review app using the Cloud<br/>Native images built by the CNG-mirror pipeline.<br><br>Cloud Native images are deployed to the `review-apps`<br>Kubernetes (GKE) cluster, in the GCP `gitlab-review-apps` project."]; D[CNG-mirror]; E[review-qa-smoke, review-qa-blocking, review-qa-non-blocking<br><br>gitlab-qa runs the e2e tests against the review app.]; A --> B1 B1 --> B B -.->|triggers a CNG-mirror pipeline| D D -.->|depends on the multi-project pipeline| B B --> C C --> E subgraph "1. gitlab-org/gitlab parent pipeline" A B1 end subgraph "2. gitlab-org/gitlab child pipeline" B C E end subgraph "CNG-mirror pipeline" D>Cloud Native images are built]; end

Detailed explanation

  1. On every pipeline during the prepare stage, the compile-production-assets job is automatically started.
  2. Once compile-production-assets is done, the review-build-cng job triggers a pipeline in the CNG-mirror project.
    • The review-build-cng job automatically starts only if your MR includes CI or frontend changes. In other cases, the job is manual.
    • The CNG-mirror pipeline creates the Docker images of each component (for example, gitlab-rails-ee, gitlab-shell, gitaly etc.) based on the commit from the GitLab pipeline and stores them in its registry.
    • We use the CNG-mirror project so that the CNG, (Cloud Native GitLab), project’s registry is not overloaded with a lot of transient Docker images.
  3. Once review-build-cng is done, the review-deploy job deploys the review app using the official GitLab Helm chart to the review-apps Kubernetes cluster on GCP.
  4. Once the review-deploy job succeeds, you should be able to use your review app thanks to the direct link to it from the MR widget. To log into the review app, see “Sign in to my review app?” below.

Additional notes:

  • If the review-deploy job keeps failing (and a manual retry didn’t help), post a message in the #g_qe_engineering_productivity channel and/or create a ~"Engineering Productivity" ~"dx::review apps" ~"type::bug" issue with a link to your merge request. The deployment failure can reveal an actual problem introduced in your merge request (that is, this isn’t necessarily a transient failure)!
  • If the review-qa-smoke job keeps failing (we already retry them once), check the job’s logs: you could discover an actual problem introduced in your merge request. You can also download the artifacts to see screenshots of the page at the time the failures occurred. If you don’t find the cause of the failure or if it seems unrelated to your change, post a message in the #test-platform channel and/or create a ~Quality ~"type::bug" issue with a link to your merge request.
  • The manual review-stop can be used to stop a review app manually, and is also started by GitLab once a merge request’s branch is deleted after being merged.
  • The Kubernetes cluster is connected to the gitlab projects using the GitLab Kubernetes integration. This basically allows to have a link to the review app directly from the merge request widget.

Auto-stopping of review apps

Review apps are automatically stopped 2 days after the last deployment thanks to the Environment auto-stop feature.

If you need your review app to stay up for a longer time, you can pin its environment or retry the review-deploy job to update the “latest deployed at” time.

The review-cleanup job that automatically runs in scheduled pipelines stops stale review apps after 5 days, deletes their environment after 6 days, and cleans up any dangling Helm releases and Kubernetes resources after 7 days.

Cluster configuration

The cluster is configured via Terraform in the engineering-productivity-infrastructure project.

Node pool image type must be Container-Optimized OS (cos), not Container-Optimized OS with Containerd (cos_containerd), due to this known issue on the Kubernetes executor for GitLab Runner

Helm

The Helm version used is defined in the registry.gitlab.com/gitlab-org/gitlab-build-images:gitlab-helm3.5-kubectl1.17 image used by the review-deploy and review-stop jobs.

Diagnosing unhealthy review app releases

If review app stability dips this may be a signal that the review-apps cluster is unhealthy. Leading indicators may be health check failures leading to restarts or majority failure for review app deployments.

The review apps Overview dashboard aids in identifying load spikes on the cluster, and if nodes are problematic or the entire cluster is trending towards unhealthy.

See the review apps page of the Engineering Productivity Runbook for troubleshooting review app releases.

Frequently Asked Questions

Isn’t it too much to trigger CNG image builds on every test run? This creates thousands of unused Docker images.

We have to start somewhere and improve later. Also, we’re using the CNG-mirror project to store these Docker images so that we can just wipe out the registry at some point, and use a new fresh, empty one.

How do we secure this from abuse? Apps are open to the world so we need to find a way to limit it to only us.

This isn’t enabled for forks.

Other resources

Helpful command-line tools

  • K9s - enables CLI dashboard across pods and enabling filtering by labels
  • Stern - enables cross pod log tailing based on label/field selectors

Return to Testing documentation