Question

Difficulty: EasyRelease Management and Deployment Strategies

An SRE team plans to execute a canary release for a critical service deployed on Google Kubernetes Engine (GKE) to safely introduce a new application revision. Order the following operational steps in the correct sequence from start to finish.

  1. 1Deploy a new revision of the workload alongside the existing production deployment.
  2. 2Configure traffic splitting on the ingress gateway to direct a minor fraction (e.g., 5%) of live user traffic to the new revision.
  3. 3Monitor telemetry logs and Service Level Indicators (SLIs) for elevated latency or error rates.
  4. 4Shift 100% of incoming production traffic to the new revision and decommission the legacy pods.

Answer

The correct sequence for conducting a canary release is to first deploy the new workload revision alongside the stable baseline, route a small fraction of traffic to it, monitor observability metrics for reliability, and finally shift 100% of traffic to the new revision while decommissioning old resources.
A standard canary release follows a structured progression: provision the new revision, split a small percentage of live traffic to test real-world behavior, evaluate telemetry against SLO performance standards, and finally promote the release to 100% traffic while terminating the old workload.

Step-by-Step Solution

1
Deploy the new container revision to GKE without routing production traffic to it yet.
The canary workload is running independently next to the production workload.
Allows verification of basic health checks before exposing any real user requests.
2
Update traffic splitting rules at the load balancer/ingress level to route 5% of traffic to the canary revision.
A controlled subset of users hits the new application code.
Limits exposure and potential business impact in case hidden defects exist.
3
Analyze error rates, latency distribution, and CPU/memory utilization using Cloud Monitoring and Logging.
Metrics confirm the canary revision meets defined SLOs without consuming excessive error budget.
Empirical verification of reliability is required before proceeding with full deployment.
4
Promote the canary revision to receive 100% of production traffic and scale down old pods.
Deployment is complete and resource utilization is optimized.
Finalizes the release lifecycle safely.

Key Concept

Canary Deployment Workflow and Traffic Splitting
Rate this question