GKE issues about crds-catalog HOT 9 CLOSED

This is just a workaround for now. In the future we will convert this utility to use the API instead.

from crds-catalog.

It works in Civo as well so I'm guessing the issue is GKE-specific.

from crds-catalog.

It looks like it's failing when trying to access the cluster to download all the CRDs manifests.
do you see any error if you're running the following command:
kubectl get crds -o yaml?

from crds-catalog.

If have full admin rights in that cluster. Retrieving CRDs with kubectl works.

from crds-catalog.

@vfarcic Hi Viktor:)
From the first line of your output, it seems like the CRD name is not parsed correctly.
Can you please run kubectl get crds and paste the output here?
Also, it would help to know which CRD failed so we can try to reproduce.
Thanks!

from crds-catalog.

Reproduced it on a fresh cluster to discard that there is something "special" in the one I used before.

Here's the list of CRDs.

kubectl get crds

W0816 19:42:16.377707   95988 gcp.go:120] WARNING: the gcp auth plugin is deprecated in v1.22+, unavailable in v1.25+; use gcloud instead.
To learn more, consult https://cloud.google.com/blog/products/containers-kubernetes/kubectl-auth-changes-in-gke
NAME                                             CREATED AT
backendconfigs.cloud.google.com                  2022-08-16T16:54:33Z
capacityrequests.internal.autoscaling.gke.io     2022-08-16T16:54:15Z
frontendconfigs.networking.gke.io                2022-08-16T16:54:34Z
managedcertificates.networking.gke.io            2022-08-16T16:54:31Z
memberships.hub.gke.io                           2022-08-16T16:57:56Z
serviceattachments.networking.gke.io             2022-08-16T16:54:35Z
servicenetworkendpointgroups.networking.gke.io   2022-08-16T16:54:35Z
updateinfos.nodemanagement.gke.io                2022-08-16T16:54:44Z
volumesnapshotclasses.snapshot.storage.k8s.io    2022-08-16T16:54:43Z
volumesnapshotcontents.snapshot.storage.k8s.io   2022-08-16T16:54:43Z
volumesnapshots.snapshot.storage.k8s.io          2022-08-16T16:54:44Z

The script might not be able to handle warnings from gcloud (the top of the output). gcloud (used by kubectl) is sending warnings to stderr.

from crds-catalog.

Yes this is definitely the issue, the script parses the output and assumes no warnings/errors.

A quick google search turned up a possible solution:
https://stackoverflow.com/a/72302509/9757520

We can try some hacky bash stuff to ignore such warnings, but solving them seems like the right way to go.

What do you think?:)

from crds-catalog.

That would fix that particular issue. Nevertheless, gcloud tends to output warnings quite often so it would be a very short-term fix. Also, there are other warnings unrelated to gcloud. For example, if there are many CRDs in a cluster, kubctl will output a warning about the delayed response. That would likely result in the same issue. You might want to query Kube API directly instead of going through kubectl and parsing the output or, if that's not an option, figure out whether something is a warning or an actual error and, if it's the former, ignore it. The exit code might help (did not check out what it gives in this case), removing the lines with WARNING, or something similar. Ultimately, parsing output is unreliable, hence we have APIs :)

from crds-catalog.

@vfarcic
We found a way to ignore the text that comes before the actual table of CRDs, so looks like this solves the problem:)
A new release is available which contains the fix. Let us know that it works for you 👍🏼

from crds-catalog.