<@U017TPA7FEW> I think there's an issue with the n...
# kubernetesw
worried-city-86458
09/30/2021, 2:08 AM@sparse-park-68967 I think there's an issue with the new helm releases, their dependencies and uninstall / delete
AFAICT, under the hood the call to helm uninstall does not wait, so dependencies do not unwind in the correct order.
When this deletes a resource with a finalizer then other resources are orphaned and it cascades from there...
s
sparse-park-68967
09/30/2021, 2:13 AMah hmm that is a possibility. We have await for install but I can see that not being readily available in delete.
sparse-park-68967
09/30/2021, 2:13 AMCan you file an issue? I will have some time to investigate tomorrow.
w
worried-city-86458
09/30/2021, 2:13 AMI'm installing helm releases using atomic, which implicitly means wait, then depend on the helm release resource elsewhere.
Since the uninstall is implicit via destroy or similar then I guess it needs to track if there are dependencies and so run with wait.
https://helm.sh/docs/helm/helm_uninstall/
https://github.com/pulumi/pulumi-kubernetes/blob/1c718d14e762ab6661bd35f9de48ab5bbdd2130e/provider/pkg/provider/helm_release.go#L985
worried-city-86458
09/30/2021, 2:14 AMSure, I'll create an issue.
s
sparse-park-68967
09/30/2021, 2:26 AMI don't think we have an easy way to see dependency information at this level in the provider but it seems reasonable to make delete wait by default...
sparse-park-68967
09/30/2021, 2:26 AMOr at least if skipAwait is false on the release resource, make uninstall await as well
w
worried-city-86458
09/30/2021, 3:17 AMworried-city-86458
09/30/2021, 3:48 AMI'm thinking the same thing, but atomic needs to be accounted for too since it implies wait.
worried-city-86458
09/30/2021, 3:49 AM(I'm currently only specifying the atomic option)
s
sparse-park-68967
09/30/2021, 5:37 AM👍
5 Views