Automation and CI/CD for Reliability
What the exam tests
This domain tests how you deliver changes repeatedly without eroding reliability. Expect infrastructure as code, IBM Cloud Schematics and Terraform, IBM Cloud Continuous Delivery pipelines, canary and blue-green releases, automated rollback, drift detection, secret handling, and change quality gates.
Core rules
- Define infrastructure as code and keep it in version control. Schematics can plan and apply Terraform for IBM Cloud resources; pipelines should run plan and require review for risky changes.
- Build one artifact and promote the same artifact through environments. Do not rebuild differently per environment.
- Use canary or blue-green deployment to limit blast radius. Define measurable promotion gates such as error rate, latency, and SLO burn.
- Automate rollback and test it. A rollback path that has never run is not a control.
- Put secrets in IBM Cloud Secrets Manager, not in Git, pipeline logs, or container images. Use service IDs and scoped IAM policies.
- Detect drift between code and deployed state. Reconcile through automation rather than manual console changes.
- Separate environments by account or resource group and apply least privilege to pipeline service IDs.
Worked example. A pipeline deploys a new image to 10% of traffic in one zone, watches 5xx rate and p95 latency for 10 minutes, and automatically rolls back if error-budget burn exceeds twice the expected rate. If the gate passes, traffic moves to 50%, then 100%. Reasoning: a small canary plus a measurable gate plus automated rollback limits exposure and shortens recovery. The gate must use the same SLI users experience, not just pod health.
Traps that cost marks
Manual hotfixes that bypass the pipeline; rollback that depends on rebuilding an old commit; secrets in environment variables or logs; infrastructure as code used only for initial creation; no drift detection; canary with no real traffic split; security and load tests omitted from the gate.
Key points
- Infrastructure as code with review and drift detection.