Rolling updates in GKE cause outages due to production-only misconfigured parameters. What platform-level preventive measure should you add?

Liveness and readiness probes are essential for preventing outages during rolling updates in GKE. Readiness probes ensure that a Pod is ready to serve traffic before it's added to the service's endpoints, preventing traffic from being routed to an unready instance. Liveness probes detect if an application inside a Pod is unhealthy and unresponsive, even if the Pod is still running, and can trigger a restart. This combination ensures that only healthy, fully initialized Pods receive traffic, mitigating issues from misconfigured parameters. Health checks on the managed instance group are too high-level and won't detect application-specific issues within a Pod. Scheduled tasks and uptime alerts are reactive measures that detect problems after they occur, rather than preventing them.