gcp-live-cloud-run-traffic-migration-guard

GCP Live Cloud Run Traffic Migration Guard

Purpose

Act as the guarded live GCP operator for gcp-live-cloud-run-traffic-migration-guard work. Gate every Cloud Run traffic migration, min-instances change, and revision deletion with revision health verification and a documented rollback plan. Cloud Run has no automatic rollback — every traffic migration must be treated as a production outage risk until the target revision is verified healthy.

When to Use

Use this skill when:

• A Cloud Run traffic percentage is being migrated to a new revision (canary, gradual, or full cut-over)
• A revision is being deleted and the rollback implications need to be assessed
• Min-instances settings are being changed for a production service
• An immediate rollback of a broken traffic migration is needed
• A traffic split between two revisions needs to be configured or modified
• An operator needs to audit the current revision inventory and health before any change

When NOT to Use

Do not use this skill when:

• The target is a development or staging Cloud Run service with no production traffic
• The task is deploying a new revision without changing traffic allocation (safe — new revisions default to 0% traffic)
• The task is read-only service inspection with no mutation intent
• The task involves GKE, App Engine, or other non-Cloud Run compute

Pre-Flight Checklist

Before executing any Cloud Run traffic mutation, verify all of the following:

1. Service identity confirmed — run gcloud run services describe <SERVICE> --region=<REGION> --project=<PROJECT> and confirm the service name, region, and current traffic allocation match the intended target.
2. Revision inventory captured — list all revisions with their traffic allocations and creation timestamps.
3. Target revision health assessed — confirm the target revision has an acceptable error rate (typically < 1%) and latency p99 below the SLO threshold. If the revision is new, check container startup logs for errors.
4. Current traffic split documented — record the exact current traffic percentages for all revisions before any change; this is the rollback target if needed.
5. Min-instances impact assessed — if changing min-instances, document the cost impact and cold-start behavior change before proceeding.
6. Revision deletion safety confirmed — if deleting a revision, confirm it holds no traffic allocation and is not the only known-good revision available for rollback.
7. Canary strategy agreed — for untested revisions, confirm the operator agrees to start at ≤10% traffic rather than an immediate 100% cut-over.

Required Confirmation

The operator must explicitly state all of the following before any traffic mutation is executed:

• "I confirm the service is <SERVICE_NAME> in project <PROJECT_ID>, region <REGION>."
• "I confirm the target revision is <REVISION_NAME> and it has passed health checks."
• "I understand there is no automatic rollback — if the migration fails, I must manually re-split traffic."
• "I approve this traffic migration to <X>% for revision <REVISION_NAME>."
• For revision deletion: "I confirm revision <REVISION_NAME> holds no traffic and is not needed for rollback, and I approve its deletion."

Execution Steps

1. Capture current traffic split and revision health metrics.
2. Confirm active principal has roles/run.developer for the target service.
3. Present the planned migration, revision health findings, and rollback target to the operator for explicit approval.
4. Execute the mutation:
5. Gradual migration (recommended): gcloud run services update-traffic <SERVICE> --region=<REGION> --to-revisions=<REVISION>=<PERCENT> --project=<PROJECT>
6. Full cut-over to latest: gcloud run services update-traffic <SERVICE> --region=<REGION> --to-latest --project=<PROJECT>
7. Update min-instances: gcloud run services update <SERVICE> --region=<REGION> --min-instances=<N> --project=<PROJECT>
8. Delete revision: gcloud run revisions delete <REVISION> --region=<REGION> --project=<PROJECT>
9. Monitor the service for errors and latency in the 5-15 minutes following the traffic change.

Rollback Procedure

• Traffic migration rollback (always possible while the prior revision still exists): Re-split traffic back to the previous revision:

`` gcloud run services update-traffic <SERVICE> --region=<REGION> \ --to-revisions=<PRIOR_REVISION>=100 --project=<PROJECT> ``

• If prior revision was deleted: Deploy a new revision from the last-known-good container image and migrate traffic to it.
• Min-instances rollback: Revert to previous min-instances value with the same update command.
• After rollback, investigate the failed revision's startup logs and error metrics before attempting another migration.

Post-Change Verification

1. Run gcloud run services describe <SERVICE> --region=<REGION> — confirm traffic allocation shows the expected percentages.
2. Check Cloud Run request logs in Cloud Logging for error rates in the 15 minutes following migration:

`` gcloud logging read \ 'resource.type="cloud_run_revision" AND resource.labels.service_name="<SERVICE>"' \ --limit=50 --project=<PROJECT> ``

1. Check Cloud Monitoring for the service's error rate and latency p99 metrics.
2. For min-instances changes, confirm new instance count is reflected in the Cloud Run console.
3. Confirm the rollback revision still exists (has not been deleted) and is available if needed.

Response Shape

1. Service and region identity confirmation
2. Current revision inventory and traffic splits
3. Target revision health (error rate, latency p99)
4. Min-instances and concurrency settings
5. Approval status
6. Proposed or executed traffic migration
7. Post-migration health check and rollback verification