should `helm_resource` always use `k8s_context()`?

[znd4]

It feels like helm_resource should respect k8s_context() by passing --kube-context={k8s_context()} (pseudocode) to helm. I've got a naive, untested implementation here, but I'm guessing there might be a better, more maintainable solution?

[nicks]

i believe the way it currently works is that tilt generates a custom kubeconfig for subprocesses to use with only one context, sets the KUBECONFIG environment variable, and helm respects that. No need to pass --kube-context directly

[nicks]

There doesn't appear to be any bug report here, so I'm going to close this issue. Let me know if I missed something!

[znd4]

Update: this definitely doesn't work as intended, at least for the delete_cmd part of helm_resource.

Can we reopen this and #598 ?

[nicks]

Sure, can you add more details on what you tried, what you expected to happen, and what happened instead?

[nicks]

for what it's worth, here are the repro steps i tried:

1. Created two kubectl contexts -- kind, and a dummy kubectl context connected to a dead cluster.
2. kubectl config use-context kind-kind
3. Ran tilt up in https://github.com/tilt-dev/tilt-extensions/tree/master/helm_resource/test
4. kubectl config use-context dummy
5. tilt trigger helloworld
6. Confirmed that the helm chart was deleted and re-installed to the kind cluster successfully

[znd4]

Instead, for 2, kubectl config use-context dummy, then 3 tilt up --context kind-kind

[znd4]

(thx for trying to repro btw :) )