GCR Cleaner

GCR Cleaner deletes untagged images in Google Cloud Container Registry or Google Cloud Artifact Registry. This can help reduce costs and keep your container images list in order.

GCR Cleaner is designed to be deployed as a Cloud Run service and invoked periodically via Cloud Scheduler.

+-------------------+    +-------------+    +-------+
|  Cloud Scheduler  | -> |  Cloud Run  | -> |  GCR  |
+-------------------+    +-------------+    +-------+

This is not an official Google product.

Setup

1. Install the Cloud SDK for your operating system. Alternatively, you can run these commands from Cloud Shell, which has the SDK and other popular tools pre-installed.

2. Export your project ID as an environment variable. The rest of this setup assumes this environment variable is set.

   export PROJECT_ID="my-project"

   Note this is your project ID, not the project number or name.

3. Enable the Google APIs - this only needs to be done once per project:

   gcloud services enable --project "${PROJECT_ID}" \
     appengine.googleapis.com \
     cloudscheduler.googleapis.com \
     run.googleapis.com

   This operation can take a few minutes, especially for recently-created projects.

4. Create a service account which will be assigned to the Cloud Run service:

   gcloud iam service-accounts create "gcr-cleaner" \
     --project "${PROJECT_ID}" \
     --display-name "gcr-cleaner"

5. Deploy the gcr-cleaner container on Cloud Run running as the service account just created:

   gcloud --quiet run deploy "gcr-cleaner" \
     --async \
     --project ${PROJECT_ID} \
     --platform "managed" \
     --service-account "gcr-cleaner@${PROJECT_ID}.iam.gserviceaccount.com" \
     --image "us-docker.pkg.dev/gcr-cleaner/gcr-cleaner/gcr-cleaner" \
     --region "us-central1" \
     --timeout "60s"

6. Grant the service account access to delete references in Google Container Registry (which stores container image laters in a Cloud Storage bucket):

   gsutil acl ch -u gcr-cleaner@${PROJECT_ID}.iam.gserviceaccount.com:W gs://artifacts.${PROJECT_ID}.appspot.com

   To cleanup refs in other GCP projects, replace PROJECT_ID with the target project ID. For example, if the Cloud Run service was running in "project-a" and you wanted to grant it permission to cleanup refs in "gcr.io/project-b/image", you would need to grant the Cloud Run service account in project-a permission on artifacts.projects-b.appspot.com.

7. Create a service account with permission to invoke the Cloud Run service:

   gcloud iam service-accounts create "gcr-cleaner-invoker" \
     --project "${PROJECT_ID}" \
     --display-name "gcr-cleaner-invoker"

   gcloud run services add-iam-policy-binding "gcr-cleaner" \
     --project "${PROJECT_ID}" \
     --platform "managed" \
     --region "us-central1" \
     --member "serviceAccount:gcr-cleaner-invoker@${PROJECT_ID}.iam.gserviceaccount.com" \
     --role "roles/run.invoker"

8. Create a Cloud Scheduler HTTP job to invoke the function every week:

   gcloud app create \
     --project "${PROJECT_ID}" \
     --region "us-central" \
     --quiet

   # Replace this with the full name of the repository for which you
   # want to cleanup old references, for example:
   export REPO="gcr.io/${PROJECT_ID}/my-image"
   export REPO="us-docker-pkg.dev/${PROJECT_ID}/my-repo/my-image"

   # Capture the URL of the Cloud Run service:
   export SERVICE_URL=$(gcloud run services describe gcr-cleaner --project "${PROJECT_ID}" --platform "managed" --region "us-central1" --format 'value(status.url)')

   gcloud scheduler jobs create http "gcrclean-myimage" \
     --project ${PROJECT_ID} \
     --description "Cleanup ${REPO}" \
     --uri "${SERVICE_URL}/http" \
     --message-body "{\"repo\":\"${REPO}\"}" \
     --oidc-service-account-email "gcr-cleaner-invoker@${PROJECT_ID}.iam.gserviceaccount.com" \
     --schedule "0 8 * * 2" \
     --time-zone="US/Eastern"

   You can create multiple Cloud Scheduler instances against the same Cloud Run service with different payloads to clean multiple GCR repositories.

9. (Optional) Run the scheduled job now:

   gcloud scheduler jobs run "gcrclean-myimage" \
     --project "${PROJECT_ID}"

   Note: for initial job deployments, you must wait a few minutes before invoking.

Payload & Parameters

The payload is expected to be JSON with the following fields:

• repo - Full name of the repository to clean, in the format gcr.io/project/repo. This field is required.

• grace - Relative duration in which to ignore references. This value is specified as a time duration value like "5s" or "3h". If set, refs newer than the duration will not be deleted. If unspecified, the default is no grace period (all untagged image refs are deleted).

• allow_tagged - If set to true, will check all images including tagged. If unspecified, the default will only delete untagged images.

• keep - If an integer is provided, it will always keep that minimum number of images. Note that it will not consider images inside the grace duration.

Running locally

In addition to the server, you can also run GCR Cleaner locally for one-off tasks using cmd/gcr-cleaner-cli:

docker run -it us-docker.pkg.dev/gcr-cleaner/gcr-cleaner/gcr-cleaner-cli

I just want the container!

You can build the container yourself using the included Dockerfile. Alternatively, you can source a pre-built container from Artifact Registry or Container Registry. All of the following URLs provide an equivalent image:

gcr.io/gcr-cleaner/gcr-cleaner
asia-docker.pkg.dev/gcr-cleaner/gcr-cleaner/gcr-cleaner
europe-docker.pkg.dev/gcr-cleaner/gcr-cleaner/gcr-cleaner
us-docker.pkg.dev/gcr-cleaner/gcr-cleaner/gcr-cleaner

FAQ

How do I clean up multiple Google Container Registry repos at once?
To clean multiple repos, create a Cloud Scheduler job for each repo, altering the payload to use the correct repo.

Does it work with Cloud Pub/Sub?
Yes! Just change the endpoint from /http to /pubsub!

What was your inspiration?
GCR Cleaner is largely inspired by ahmetb's gcrgc.sh, but it is written in Go and is designed to be run as a service.

License

This library is licensed under Apache 2.0. Full license text is available in LICENSE.