Deploy

Deployment

Pre-Deploy Checklist

• [ ] All tests pass on the branch being deployed
• [ ] Code reviewed and approved
• [ ] No known critical bugs in the diff
• [ ] Database migrations tested (if any)
• [ ] Environment variables / secrets configured for target environment
• [ ] Monitoring and alerting in place
• [ ] Rollback plan documented
• [ ] Team notified of deployment

Deployment Strategies

Rolling Deploy

Replace instances one at a time. Old and new versions run simultaneously during rollout.

• Pro: Zero downtime, simple
• Con: Must handle two versions running at once (API compatibility, DB schema)
• Use when: Changes are backward-compatible

Blue-Green Deploy

Run two identical environments. Switch traffic from blue (old) to green (new) all at once.

• Pro: Instant rollback (switch back to blue), clean cut
• Con: Requires double infrastructure during deploy
• Use when: You need guaranteed instant rollback

Canary Deploy

Route a small percentage of traffic (1-5%) to the new version. Gradually increase if healthy.

• Pro: Limits blast radius, real production validation
• Con: More complex routing, must monitor per-version metrics
• Use when: High-risk changes, large user base

Recreate

Stop old version, start new version. Simplest possible deploy.

• Pro: No version compatibility issues, simple
• Con: Downtime during switchover
• Use when: Maintenance window is acceptable, internal tools

Deploy Steps

1. PREPARE
   - Build production artifact
   - Run smoke tests against artifact
   - Verify configuration for target environment

2. DEPLOY
   - Apply database migrations (if any, BEFORE new code)
   - Deploy new code using chosen strategy
   - Monitor for errors during rollout

3. VERIFY
   - Health check endpoints responding
   - Key user flows working (smoke test)
   - Error rate not elevated
   - Latency within normal range
   - No new error types in logs

4. CONFIRM or ROLLBACK
   - If healthy: mark deploy as successful
   - If unhealthy: execute rollback plan immediately

Rollback

When to Rollback

• Error rate spikes above baseline
• Health checks failing
• Key user flows broken
• Performance degradation beyond acceptable threshold

How to Rollback

1. Deploy the previous known-good version (don't try to fix forward under pressure)
2. If database migrations were applied: run the down migration FIRST, then rollback code
3. Verify service is healthy after rollback
4. Investigate root cause at leisure, not under incident pressure

Migrations and Rollback

• Forward-compatible migrations: New columns nullable, old columns still present
• Never drop columns in the same deploy as the code change
• Backfill data in a separate step from schema change
• If a migration can't be rolled back: mark this in the deploy plan and increase monitoring

Post-Deploy

• [ ] Verify key flows working on production
• [ ] Check error monitoring for new errors
• [ ] Check performance metrics
• [ ] Update deployment log / changelog
• [ ] Close related tickets
• [ ] Monitor for 15-30 min before moving on

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: pvnarp
• Source: pvnarp/agent-skills
• License: MIT

Install and usage instructions live in the source repository linked above.