Anyone seeing this error after the recent update? ...
# kubernetes
k
Anyone seeing this error after the recent update?
Copy code
error: an unhandled error occurred: program failed:
    waiting for RPCs: failed to decode YAML: rpc error: code = Unknown desc = invocation of kubernetes:yaml:decode returned an error: failed to initialize discovery client: The gcp auth plugin has been removed.
    Please use the "gke-gcloud-auth-plugin" kubectl/client-go credential plugin instead.
    See <https://cloud.google.com/blog/products/containers-kubernetes/kubectl-auth-changes-in-gke> for further details
Have checked the URL and tried setting the USE_GKE_GCLOUD_AUTH_PLUGIN=True environment variable with no luck. Have updated all other dependencies that I can find as well as gcloud and kubectl.
d
Your kubeconfig may need updating to use the new auth plugin
k
I nuked the kube config (it was old from something else and not intended to be being used and that resolved the issue. Odd becuase I didn’t want to use that config at all. Didn’t realize it waas using whatever my default tehre was. I would expect it to override it.
d
if you specified one, it should definitely overwrite. Are you using an explicit provider? perhaps it fell back to using the default provider somewhere
k
I’m using an explicit provider which I would have imagined would override my local config. I shouldn’t have to nuke my local config to use a pulumi created provider.
d
It's worth adding this to your config, to prevent pulumi accidentally using the default providers: https://www.pulumi.com/docs/concepts/config/#pulumidisable-default-providers
k
Oh good call, that will help!
The repo you linked is helpful too. Great to see how others are actually putting this in code.
It seems even with disable default providers for kubernetes, I still have to move my ~/.kube/config file to some other name in order for the Kubeconfig: [mykubeconfigstring] param to take any effect. Is this a known issue? Or maybe I’m missing something on how to set it up. The code is here if it helps: https://github.com/phutchins/dimo-node Specifically the kubeProvider: https://github.com/phutchins/dimo-node/blob/main/dependencies/dependencies.go#L26
d
k
No, that one looks like it’s working just fine. Which is why I’m confused. 🙂
d
I'm not too familiar with golang. Perhaps if BuildInfrastructurer returned the kubeconfig directly, it may work?
d
@kind-fireman-33438 you might be hitting a problem related to how the
provider
option propagates thru component resources like
Chart
. Note that the behavior varies across the SDKs to some extent. https://github.com/pulumi/pulumi-kubernetes/issues/2254 I'm working on a fix, please stay tuned (and feel free to test it out if you're so inclined). https://github.com/pulumi/pulumi-kubernetes/pull/2624
d
Ohai, @damp-airline-38442. I'm antdking :) It looks like the provider is specified explicitly in this case, though maybe the go-sdk is less clear in explicit vs multiple providers with its DSL. In any case, thanks for looking into the issue
k
Hey @damp-airline-38442, thanks for the update. I’ll definitely keep an eye on that issue and PR. Could it be that if you specify a provider that was given an empty string for its kubeconfig, it reverts to the default config/provider?
d
I would tentatively say no that it would not fall back to a default provider. More likely that the provider isn't propagating. That said when a provider is misconfigured, it falls back to a degraded mode of behavior.