The Pod suddenly kept restarting, with abnormal traffic coming in.
ReadinessProbe
and LivenessProbe
were configured, it was highly likely that all health checks failed more times than the upper limit set in LivenessProbe
, thereby leading to a restart.preStop
for graceful termination, indicating to run preStop
before restart. The graceful termination took a long time, and during preStop
execution, ReadinessProbe
kept probing.ReadinessProbe
but not LivenessProbe
determines whether a Pod is ready. As ReadinessProbe
was successful during preStop
execution, the Pod became ready.initialDelaySeconds
or StartProbe to protect containers that start slowly.ReadinessProbe
succeeds and lets traffic in, which will not be handled by the business, leading to traffic exceptions. We recommend you use a better probe method, where the business provides the HTTP liveness probe to check the actual health status of the business.
Was this page helpful?