Lewati ke konten utama

Runbook — Deploy dashboard-api

Steps to deploy the dashboard-api service to the staging / production environment.

Prerequisites​

  • Access to the target server (SSH key at ~/.ssh/kesles-prod)
  • KESLES_SECRET_ENCRYPTION_KEY registered in the server env
  • Target PostgreSQL has been migrated to the latest version (see db-schema)

Deploy Steps​

1. Build binary locally​

cd services/dashboard_api
go build -o dashboard-api ./cmd/server

2. Local smoke test​

./dashboard-api --config config-staging.yaml &
curl -fsS http://localhost:8082/health
kill %1

3. Push to server​

scp dashboard-api kesles-prod:/opt/kesles/dashboard-api/bin/
ssh kesles-prod "systemctl restart dashboard-api"

4. Verify post-deploy​

curl -fsS https://api-merchant-staging.kesles.com/health
# expect 200 {"status":"ok","build":"<git-sha>"}

5. Monitor for 10 minutes​

  • Tail log: ssh kesles-prod "journalctl -u dashboard-api -f"
  • Watch metrics dashboard: https://grafana.kesles.internal/d/dashboard-api

Rollback​

If the health check fails or error rate rises > 1%:

ssh kesles-prod "cp /opt/kesles/dashboard-api/bin/dashboard-api.prev /opt/kesles/dashboard-api/bin/dashboard-api && systemctl restart dashboard-api"

The previous binary version is always backed up to .prev on each deploy.

Production Deploy Checklist​

  • Staging has been running ≥ 24 hours without an error spike
  • Production DB migration has been executed via scripts/apply_migration.sh (tracker-aware sejak mig 140) — verify no drift: diff ls migrations/ vs SELECT name FROM public.schema_migrations. Reference: merchant_docs/docs/database/operations/migration-workflow.md.
  • PR merged to main with reviewer approval
  • Pre-prod smoke test passes
  • Oncall standby notified via Slack #oncall-engineering
  • Deploy window: Monday–Thursday, 10:00–16:00 WIB (avoid weekends/end of month)