A new container version for GKE wasn’t fully load-tested. You need an automated deployment that minimizes production performance risk. What should you do?

Canary deployments are ideal for minimizing risk with untested versions. They allow you to gradually roll out the new version to a small subset of users, monitor its performance with Cloud Monitoring, and then incrementally increase traffic if no issues are detected. This limits the blast radius of any potential problems. Blue/green deployments switch all traffic at once, which is riskier for untested code. Rolling updates with kubectl are a basic deployment strategy but lack the fine-grained traffic control and automated promotion/rollback capabilities of a dedicated canary deployment system within a CI/CD pipeline. Cloud Load Balancing can manage traffic, but a full CI/CD pipeline with canary deployment capabilities provides a more integrated and automated solution for risk mitigation.