Use auto-build-image for build stage
package-and-qajob from pipeline https://gitlab.com/gitlab-org/gitlab-ce/pipelines/48987002 triggered https://gitlab.com/gitlab-org/omnibus-gitlab/pipelines/48988994 downstream.
Trigger:qa-testjob from pipeline https://gitlab.com/gitlab-org/omnibus-gitlab/pipelines/48988994 triggered https://gitlab.com/gitlab-org/gitlab-qa/pipelines/48992005 downstream.
Updated my .gitlab-ci.yml based on the latest from yesterday which included this (along with a couple more updates)
With a node detected app, 95% of the time the liveness/health checks will just fail with no logs whatsoever like my
startscript was never getting called. Created a basic hello-world application which worked and slowly increased it in complexity to try and see what was causing my problems deploying.
Eventually I ran into the deployment issue and none of my apps were deploying to staging correctly. After using kubectl on the CrashLoopBackOff pods you'll get no logs and describe will pretty much say:
Warning Unhealthy 15m (x134 over 2h) kubelet, gke-my-cluster-default-pool-8b037576-jqp8 Liveness probe failed: Get http://10.52.0.79:5000/: dial tcp 10.52.0.79:5000: connect: connection refused Warning BackOff 21s (x530 over 2h) kubelet, gke-my-cluster-default-pool-8b037576-jqp8 Back-off restarting failed containe
initialDelaySecondsof health/liveness caused me to think it was fixed but it only improved things but it was still flakey as hell.
Not sure where else to mention this but it's caused me no end of grief these past 2 days :) Reverted the above changes and my apps consistently deploy now.
@intellix1 it would be good if you could raise an issue for this so we can keep track.
The container images built with this extracted auto-build-image job should be identical to before. So I would be curious to see what the differences are in the resulting images.
Otherwise, do you have the full diff in the config? Just in case there are other changes that may affect this.