Atlas: Readiness Probe Failing

What this usually means

Your workload is running, but Kubernetes will not route traffic to it. Readiness is a gating signal; when it is wrong, it creates outages that look like networking problems.

Likely causes

Probe failures are often mis-specified semantics, not platform failure.

• Probe points at the wrong path/port or assumes localhost vs service address.
• Startup time exceeds probe thresholds; timeouts too small under load.
• Probe gates on a dependency that is allowed to be temporarily absent (causes self-inflicted unavailability).
• NetworkPolicy blocks the probe path if it crosses namespaces or uses sidecars.

What to inspect first

Read events and verify the probe endpoint from within the pod namespace.

• Observe failure message and cadence.
• Confirm the application actually listens where the probe checks.

kubectl describe pod <pod> -n <ns>
kubectl get pod <pod> -n <ns> -o yaml | rg -n "readinessProbe|httpGet|tcpSocket|exec"

Resolution guidance

Readiness should mean “can serve”. Keep it fast, meaningful, and aligned to reality.

• Adjust timeouts/failureThreshold to match real startup characteristics.
• Avoid readiness that depends on fragile external systems unless required.
• Prefer startupProbe for slow initialization and keep liveness conservative.