Make outer health probes outlast inner request deadlines

When a Kubernetes exec or HTTP probe invokes a health client with its own timeout, the probe's timeoutSeconds must exceed the client's complete deadline plus measured scheduling overhead. If the outer probe expires first—especially when timeoutSeconds is omitted and defaults low—Kubelet can kill a healthy lifecycle helper before its client returns the intended degraded result. Set the inner request deadline first, give the outer probe a margin, and test slow and failure paths as well as success. A wider probe budget prevents false restarts; it does not make a genuinely slow dependency healthy or replace threshold and cadence tuning.

Applies when: A Kubernetes probe wraps code that performs bounded network or provider checks. Keywords: Kubernetes probe timeoutSeconds Kubelet health client deadline restart readiness