Make deploy failes for Kubernetes v1.21.2 about flink-on-k8s-operator HOT 9 CLOSED

I have encountered a similar error as well

Error: unable to parse option "crd:maxDescLen=0,trivialVersions=true,generateEmbeddedObjectMeta=true": [unknown argument "generateEmbeddedObjectMeta" (at <input>:1:61) extra arguments provided: "true" (at <input>:1:62)]

from flink-on-k8s-operator.

hi @numbnut I haven't tested yet with v1.21.2 but I'll take a look.

Using [email protected] was needed before to be able to run the operator with k8s >1.18 but that has been addressed in this fork at least up until v1.20.x`

The newer0.6.1 introduces generateEmbeddedObjectMeta which is required specifically to add the metadata for the PersistentVolumeClaims without it claims will not be deleted when the cluster is deleted;

@clouddra check if your env is not using a previous version of controller-tools

from flink-on-k8s-operator.

Hi @regadas. Using the right version of controller tools worked for me.

However I did encounter another problem when deploying to v1.15.7.

kubectl kustomize config/crd > config/crd/bases/patched_crd.yaml
mv config/crd/bases/patched_crd.yaml config/crd/bases/flinkoperator.k8s.io_flinkclusters.yaml
go mod tidy
kubectl apply -f config/crd/bases
error: unable to recognize "config/crd/bases/flinkoperator.k8s.io_flinkclusters.yaml": no matches for kind "CustomResourceDefinition" in version "apiextensions.k8s.io/v1"

The issue is likely due to the CRD's apiVersion. It worked after I modified manually the generated yaml manifest to match that of apiextensions.k8s.io/v1beta1. For anyone interested please refer to the public gist for the changes made.

from flink-on-k8s-operator.

Hi @regadas , thanks for the feedback. If I can somehow support you, please let me know.

from flink-on-k8s-operator.

Hi @clouddra , I'm not sure to which Kubernetes version you are referring. Are all your comments referring to the version v1.15.7, which you are mentioning in the second comment? If this is the case I think it might be better to open a new ticket, because this one is focusing on v1.21.x.

from flink-on-k8s-operator.

I doubled checked the operator deployment on k8s >=1.18 and it seems that there are no issues;

I used our latest version:

make deploy IMG=ghcr.io/spotify/flink-operator:v0.1.13

Make sure controller-tools is on 0.6.1

from flink-on-k8s-operator.

Hi @regadas , thanks for checking it. I will take another look and try out your pre-built image. If I have some results I will report back.

from flink-on-k8s-operator.

Hi @regadas ,
I can confirm that you prebuilt image is working on Docker for Mac. It seems that there are problems in the build chain on Docker for Mac or that I have misconfigured something.
Can you say something about MacOS compatibility?
Should I maybe create a new ticket about it?
Regards,
Niklas

from flink-on-k8s-operator.

I checked again with the latest release 0.1.14 and the problem vanished. Since a lot of changes have been applied to the build logic I assume that this might have fixed the problem.

from flink-on-k8s-operator.