Open danjenkins opened 10 months ago
I just redeploy an instance to check for this, and it runs without problem from initialisation until meeting creations in the app, without rebooting. However, I also encountered some probes failures, only at the startup of the app. using calcom-stack-0.1.2:
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Normal Pulling 15m kubelet Pulling image "calcom/cal.com:v3.5.4"
Normal Pulled 14m kubelet Successfully pulled image "calcom/cal.com:v3.5.4" in 1m8.501s (1m8.501s including waiting)
Normal Created 14m kubelet Created container calcom
Normal Started 14m kubelet Started container calcom
Warning Unhealthy 13m (x3 over 14m) kubelet Readiness probe failed: Get "http://10.0.174.18:3000/": dial tcp 10.0.174.18:3000: connect: connection refused
Warning Unhealthy 13m kubelet Liveness probe failed: Get "http://10.0.174.18:3000/": dial tcp 10.0.174.18:3000: connect: connection refused
From my point of view, there's a problem: the delay for the app to start is higher (especially at first startup with database initialisation) and can be over 2 minutes, according to my logs. I'll investigate on adding a startup probe.
For your scenario, I don't understand why it takes so much time to start. Can you send the pod logs for further investigation? Also, how much resources did you allow for the app?
I'll get back to you with some logs, I went and removed the probes to make things come up fine to get past the issue temporarily.
Resources - I didn't limit what it had available I think. Will confirm
@danjenkins any updates about your performance issue?
Sorry due to illness/christmas I havent gotten around to looking at this again. Will do soon!
@danjenkins I'll close this issue for now. Feel free to comment it with fresh updates, I'll be happy to re-open it to investigate.
No worries - I'm going to be moving my k8s cluster in the next week or so so I'll test these changes then!
@Pyrrha Do you think it will be possible to add the startup probe? 30s doesn't seem to be enough to finish the startup of the application even without resource limitation applied.
Thank you.
Hello @erueda1, As said to @danjenkins, it looks weird to me that the app startup in a so long time. Could you provide me a benchmark presenting mean time for startup, and a summary of your infrastructure? I'll conduct same on my side too. This probe will be implemented soon, next feature release I think.
I was able to briefly see the admin setup screen for cal.com self hosted but then the liveness/readiness checks failed.
But I could definitely get to the UI via the ingress before the container was killed.
I see you can't disable the liveness/readiness checks in the chart. Maybe that should be an option?
Not sure why they're failing yet.