Best value for readiness probe timeout? (kubernetes)

Grafana Configuration
1

Hello

what is the best value for readiness probe timeout?
1 second (default) => probe fails after deployment scale down & up (for backup)
5 seconds => seems OK
15 seconds => probe fails

I use grafana helm chart 7.2.5 with openshift 4.12

BR
Pascal

2

It depends, e. g. what pod is doing when it’s starting (dashboard, plugin, alert provisioning,… you should see that from pod logs and then you can estimate time just for start). I would configure initialDelaySeconds (e. g. 30sec, see estimation from previous step if you need exact value for your case) so there is a time for start and then 10sec period.

3

Hello
according to the grafana container logs, it takes 20 seconds for startup,
So I set readiness-probe-initial- delay= 30 seconds
=> it works ! thanks!!
(except 1 readiness probe failed from kubelet 1 hour after restart)

would you also suggest increase readiness-probe-timeout ? (as I did increase it from 1 to 5 seconds, and it did fix the problem also)

thanks again
Pascal

4

That depends on your case. If you have single pod, which is under heavy load then it makes sense to increase a timeout. But if you have (auto) scaled pods, which are just sitting and waiting for requests, then it doesn’t make sense - healtcheck response should be quick in this case.