Imported from packit/ai-workflows (.claude/skills/verify-prod-deployment/SKILL.md). Install upstream with npx skills add packit/ai-workflows --skill verify-prod-deployment. Copyright stays with the author.

Verify Production Deployment

Overview

Check the health of the jotnar-ymir OpenShift cluster after a fresh deployment. Run these commands in order and report findings.

Steps

1. What changed

git log --oneline -25

2. Pod status

oc get pods -n jotnar-ymir--jotnar-ymir

Expected: All agent/service pods Running, completed jobs Completed, zero Error or CrashLoopBackOff or ImagePullBackOff.

3. Deployment rollout status

oc get deploy -n jotnar-ymir--jotnar-ymir

Expected: All deployments show READY count matches desired (e.g. 2/2, 1/1).

4. Imagestreams

oc get imagestream -n jotnar-ymir--jotnar-ymir

Note the UPDATED timestamp for each imagestream. In the report, call out any agent or mcp-server imagestream updated within the last 24 hours — these are the ones that just rolled out.

5. Recent events

oc get events -n jotnar-ymir--jotnar-ymir --sort-by='.lastTimestamp' | tail -40

Look for Warning severity lines — Normal is noise. Flag anything other than routine Pulling/Pulled/Created/Started/Completed.

Reporting

Report as a short table or bullet list:

• All pods running: yes/no
• Any restarts: yes (pod name + count) / no
• Any warnings in events: yes (what) / no
• Deployments all READY: yes/no
• Imagestreams updated <24h: list each (name, tag, updated) — highlight agents and mcp-server
• Overall verdict: OK or ISSUE:

Known gotchas

• Recreate strategy: Deployments use Recreate (not RollingUpdate). During rollout there is a brief gap with 0 pods — normal, not a failure.
• Quota deadlock: If new pods show FailedCreate + quota error, check oc get appliedclusterresourcequotas. Usually resolves itself with Recreate strategy.
• jira-issue-fetcher jobs: These are cronjobs that run every 5 minutes and complete quickly — Completed status is expected and healthy.