Zookeeper readiness probes fail occasionally

[solsson]

This is a follow-up on #63. A couple of times per day the probe fails on some zookeeper pods.

These failing ruok calls are the reason we did 1f4321e.

As 2c4b6cd explains I found this sort of probe in a different zookeeper k8s setup. Maybe it's not reliable enough, or maybe zookeeper is actually unstable.

If we solve this we can reinstate the liveness probes.

[solsson]

Is there a more lightweight way (than with /bin/sh and nc) to do this probe? Or would a TCP probe on 2181 be sufficient?

[solsson]

We could test with the exact bash script from https://kubernetes.io/docs/tutorials/stateful-application/zookeeper/#testing-for-liveness