Rollbacks & Auto-Recovery

Doku automatically rolls back deployments when it detects problems. You can also trigger manual rollbacks at any time.

Automatic Rollback

During a canary rollout, Doku continuously monitors your application. A rollback is triggered automatically when any of the following thresholds are exceeded:

Error rate exceeds canary.error_threshold (default: 1%)
P99 latency exceeds canary.latency_threshold (default: 1000ms)
Health checks fail on the new version
OOM kills or crash loops detected

bash

# [canary] Step 2/4: routing 25% traffic → v42
# [canary] ✗ Error rate 3.2% exceeds threshold 1%
# [rollback] Rolling back to v41...
# [rollback] ✓ Rollback complete. v41 serving 100% traffic.

Customize rollback thresholds in your doku.yaml under the canary section. See Canary Configuration.

Manual Rollback

Roll back to the previous stable version at any time:

bash

doku rollback --env production

Or roll back to a specific deployment by ID:

bash

doku rollback --env production --to deploy_abc123

What Happens During a Rollback

Traffic is immediately shifted back to the previous version (100%)
New version containers are stopped and drained
Infrastructure changes (if any) are reverted
A notification is sent to your configured channels (Slack, webhooks, etc.)
The failed deployment is logged with metrics and root cause analysis

Zero-downtime

Rollbacks are zero-downtime. The previous version is kept warm and ready to serve traffic at all times during a deployment.

Deployment History

View past deployments and their status:

bash

doku status --env production

# Deployment History:
# deploy_def456  v42  rolled-back  2 hours ago   error rate 3.2%
# deploy_abc123  v41  active       3 hours ago   healthy
# deploy_xyz789  v40  superseded   1 day ago     healthy

Environments

CI/CD Integration