Reported failures from https://k8s-testgrid.appspot.com/kubernetes-apps
Cockroachdb:
W0613 14:19:20.823] Error: release wzq failed: StatefulSet.apps "wzq-cockroachdb" is invalid: spec.template.annotations.affinity.podAntiAffinity.preferredDuringSchedulingIgnoredDuringExecution[0].podAffinityTerm.topologyKey: Required value: can not be empty
Spinnaker:
W0613 14:19:49.774] Error: found in requirements.yaml, but missing in charts/ directory: redis, minio
Jenkins:
W0613 14:19:51.019] Error: error validating "": error validating data: unexpected type: types.UnixUserID
cc @kubernetes/charts-maintainers @viglesiasce
Jenkins issue is now resolved because upstream k8s reverted the backward incompatible API change. The test is now passing. Cockroachdb will pass after we merge https://github.com/kubernetes/charts/issues/1306 on the day 1.7 releases.
Wound it be useful if one could change an arbitrary needed field in the resources definitions (that can not be changed via values abstraction), not by cloning/copy/pasting entire chart code and modifying the template?
I think helm charts has a kind of flaw abstraction model, that doesn't allow to reuse them in more flexible way.
I'd like to add that concourse.ci is also victim to this.
What exactly is the yaml workaround btw?
What's the fix for this?
Hello this issue persists confirmed installing according to README/doc.
helm install stable/concourse --namespace concourse
Error: release funky-cricket failed: StatefulSet.apps "funky-cricket-worker" is invalid: spec.template.annotations.affinity.podAntiAffinity.preferredDuringSchedulingIgnoredDuringExecution[0].podAffinityTerm.topologyKey: Required value: can not be empty
helm status funky-cricket | head -3
LAST DEPLOYED: Fri Oct 13 14:51:52 2017
NAMESPACE: concourse
STATUS: FAILED
Note that when deploying from source repo after checking for and installing any missing dependencies:
helm dep list
helm dep update
helm dep build
. . . then helm install seems to get past the error and the deployment appears to be successful according to helm, but note that the postgresql and worker pods remain in pending state:
helm install stable/concourse -f $GIT_HOME/charts/stable/concourse/values.yaml --namespace concourse
helm status concourse | head -3
LAST DEPLOYED: Fri Oct 13 15:07:52 2017
NAMESPACE: concourse
STATUS: DEPLOYED
kubectl get all --namespace concourse
NAME DESIRED CURRENT UP-TO-DATE AVAILABLE AGE
deploy/concourse-postgresql 1 1 1 0 3m
deploy/concourse-web 1 1 1 0 3m
NAME DESIRED CURRENT READY AGE
rs/concourse-postgresql-4126483298 1 1 0 3m
rs/concourse-web-1449843452 1 1 0 3m
NAME DESIRED CURRENT UP-TO-DATE AVAILABLE AGE
deploy/concourse-postgresql 1 1 1 0 3m
deploy/concourse-web 1 1 1 0 3m
NAME DESIRED CURRENT AGE
statefulsets/concourse-worker 2 1 3m
NAME READY STATUS RESTARTS AGE
po/concourse-postgresql-4126483298-49w1v 0/1 Pending 0 3m
po/concourse-web-1449843452-1drm9 0/1 Running 1 3m
po/concourse-worker-0 0/1 Pending 0 3m
Thanks
I ran into a similar issue with stable/concourse. I had to run helm repo update to ensure that the newest chart version was getting installed. After that it worked for me.
This is likely out of date since 1.8 came out already. Individual issues should be filed for each of these against chart maintainers. Closing
Most helpful comment
I'd like to add that concourse.ci is also victim to this.
What exactly is the yaml workaround btw?