Code
Audit your Kubernetes manifests
Missing limits, bad probes, and workloads sized from a guess.
Fill it in
Where they live.
If you can point at metrics.
Root containers, host mounts, and privilege.
Your prompt
Audit the manifests in [the manifests]. Check the things that cause 3am pages: missing resource requests (so the scheduler guesses), missing limits (so one pod takes the node), liveness probes that restart a slow-starting container forever, readiness probes that let traffic in early, and a single replica with no PodDisruptionBudget. A liveness probe pointing at the same endpoint as readiness is the most common one and it turns a slow dependency into a restart loop. Rightsize from actual usage rather than from the number somebody typed, and say what you based it on. Check the security context too: containers running as root, host path mounts, privileged, and service accounts with more than they use.
Use Audit your Kubernetes manifestsOpens with everything above already filled in.
Why this works
A liveness probe pointing at the same endpoint as readiness turns a slow dependency into a restart loop, and it is the most common manifest mistake there is. This checks the things that cause pages: requests, limits, probes, replicas and disruption budgets.