Redis BUG: Redis Sentinel (7.2.4) liveness or readiness probe timeout

We’re encountering an issue with Redis in Sentinel mode. Everything works as expected, but the Redis and Sentinel containers in the pod are reporting timeouts on probes. Fortunately, the containers are not restarting. Somebody also reported similar issue in the https://github.com/bitnami/charts/issues/29985

If I exec into the pod and run some of the probes directly (e.g. /health/redis_liveness.sh) everything works as expected and response is PONG and RC is correctly 0.

Events:
  Type     Reason     Age                      From     Message
  ----     ------     ----                     ----     -------
  Warning  Unhealthy  37m (x619 over 2d18h)    kubelet  Liveness probe failed: command "sh -c /health/redis_liveness.sh" timed out
  Warning  Unhealthy  22m (x638 over 2d18h)    kubelet  Liveness probe failed: command "sh -c /health/sentinel_liveness.sh" timed out
  Warning  Unhealthy  12m (x664 over 2d18h)    kubelet  Readiness probe failed: command "sh -c /health/redis_readiness.sh" timed out
  Warning  Unhealthy  2m48s (x652 over 2d18h)  kubelet  Readiness probe failed: command "sh -c /health/sentinel_liveness.sh" timed out

        livenessProbe:
          exec:
            command:
            - sh
            - -c
            - /health/redis_liveness.sh
          failureThreshold: 5
          initialDelaySeconds: 30
          periodSeconds: 20
          successThreshold: 1
          timeoutSeconds: 40

        readinessProbe:
          exec:
            command:
            - sh
            - -c
            - /health/redis_readiness.sh
          failureThreshold: 5
          initialDelaySeconds: 30
          periodSeconds: 20
          successThreshold: 1
          timeoutSeconds: 40

To reproduce

Run Sentinel mode of Redis i 3 replicas.

Expected behavior

No timeouts on probes.

Additional information Kubernetes 1.30.5 OS: Ubuntu 22.04.5 LTS
Kernel: 5.15.0-122-generic
CNI: Cilium 1.16.2 containerd://1.6.30

Comment From: lukastopiarz

Another finding: If you exec into the container and run the same liveness/readiness probe shell script in an infinite loop, it never fails or timeouts.