Hey all, I'm receiving error messages, when I try to perform some rather simple changes of a yaml.ConfigFile. In my case, I am issuing two ClusterIssuers for cert-manager. One uses the acme staging servers, the other uses the acme prod servers. I had a bug in the prod server cluster issuer, it used the staging server. So I updated the prod ClusterIssuer to use the acme prod server and tried to deploy the change using "pulumi up". Pulumi correctly identifies the "spec" of the ClusterIssuer to be different. Nevertheless, when I let pulumi run the upgrade, I get an error message "the Kubernetes API server reported that "name-of-clusterissuer" failed to fully initialize or become live: the body of the request was in an unknown format - accepted media types include: application/json-patch+json, application/merge-patch+json, application/apply-patch+yaml". Any ideas what's going on here? Updating from changed yaml files is and will be an important use case, since not every k8s resource will (and should) be part of pulumi sdk. Can someone shed some light on this please.

You should be able to (1) comment out the resource (2) run pulumi up (3) uncomment (4) and deploy the updated yaml as a workaround

That sounds like a bug in the k8s provider. Can you file an issue with repro steps if possible? Generally, changes to the YAML file should work as you were expecting.

@rhythmic-hair-33677 Thanks mate, that worked. I'll keep that kind of workaround in my head.

Yep, thanks!

@gorgeous-egg-16927 I tried to create a minimal repro today, but I fail to reproduce the issue. I tried with a simple nginx deployment, but changing the specs there works as expected. Since cert-manager uses its own CRDs, it may be related to those. I will see if I can create a repro for that in the next days. Will let you know then.

Hello @gorgeous-egg-16927, was there any fix for this? I’m having the same issue with many Custom Resources. I start seeing it after upgrading my EKS clusters to k8s 1.15.

No, not to my knowledge. I don’t think we ever had a repro case to investigate

Does it help if I send you a verbose log of the issue happening?

That’d be great, thanks

I’ll do it in private. Thanks.