antctl check cluster tries to pull wrong tag

[mkaring]

Describe the bug

The command antctl check cluster using the version deployed with antrea v2.1.0 fails to startup the cluster-checker. The reason is that the image antrea/antrea-agent-ubuntu is requested with the tag "2.1.0" instead of "v2.1.0". This causes the image pull to fail.

  Normal   Pulling    52s (x4 over 2m24s)  kubelet            Pulling image "antrea/antrea-agent-ubuntu:2.1.0"
  Warning  Failed     51s (x4 over 2m23s)  kubelet            Failed to pull image "antrea/antrea-agent-ubuntu:2.1.0": rpc error: code = NotFound desc = failed to pull and unpack image "docker.io/antrea/antrea-agent-ubuntu:2.1.0": failed to resolve reference "docker.io/antrea/antrea-agent-ubuntu:2.1.0": docker.io/antrea/antrea-agent-ubuntu:2.1.0: not found
  Warning  Failed     51s (x4 over 2m23s)  kubelet            Error: ErrImagePull
  Warning  Failed     41s (x6 over 2m23s)  kubelet            Error: ImagePullBackOff
  Normal   BackOff    30s (x7 over 2m23s)  kubelet            Back-off pulling image "antrea/antrea-agent-ubuntu:2.1.0"

The correct image has to tag "v2.1.0" according to DockerHub: antrea/antrea-agent-ubuntu:v2.1.0

Subsequently this causes the command itself to also fail.

[kubernetes] Creating Namespace antrea-test-5uat0 for pre installation tests...
[kubernetes] Creating Deployment
[kubernetes] Waiting for Deployment to become ready
[kubernetes] Waiting for Deployment cluster-checker to become ready...
Error: error while waiting for Deployment to become ready: waiting for Deployment cluster-checker to become ready has been interrupted: context deadline exceeded

To Reproduce

1. Get the antctl executable from the v2.1.0 release.
2. Run antctl check cluster
3. Wait for the timeout error

After that you can easily inspect the pod created and check the events.

Expected

The command starts its checker pod up correctly.

Versions:

• Antrea: v2.1.0
• Kubernetes:
  • Client Version: v1.29.2
  • Kustomize Version: v5.0.4-0.20230601165947-6ce0bf390ce3
  • Server Version: v1.29.7
• Container Runtime: containerd 1.7.19
• Control Plane OS: Ubuntu 22.04.4 LTS

[luolanzone]

Hi @mkaring may I know how did you deploy the cluster with Antrea? This tool antctl check cluster is used before Antrea is installed. But according to your description, seems the Antrea is already installed but not succeeded.

[tnqn]

Thanks @mkaring for reporting it.

@luolanzone I can reproduce the issue. It's a valid one. There is an issue when getting the image tag when antctl is built for release, as @mkaring has pointed out.

[mkaring]

@luolanzone The installation of antrea did indeed fail at the point that the OVS from my windows node did not seem to be able to connect with the one on the control plane.

The check of the cluster installation also showed timeout problems, because the containers did not start up. The reason for that is that the antrea agent did not start on my windows node.

As I was stuck at this point, I had the idea to tear down the entire cluster and start again from scratch, to see what the checker has to say about it.

[luolanzone]

Hi @mkaring , thanks for the info, we confirmed that it is indeed a problem in antctl check cluster, @tnqn had submitted a bugfix via PR #6565.