Running `kind-setup.sh` with no arguments fails to locate image

[chipzoller]

Describe the bug

When following the instructions given here and running kind-setup.sh with no arguments (on KinD v0.14), the script is unable to locate and therefore load the image named projects.registry.vmware.com/antrea/antrea-ubuntu:latest.

To Reproduce

1. Have KinD v0.14 installed.
2. Clone the antrea repo locally.
3. Run ./ci/kind/kind-setup.sh create antrea per documentation instructions.
4. See the following output printed (last line)

No kind clusters found.
Creating cluster "antrea" ...
 ✓ Ensuring node image (kindest/node:v1.24.0) 🖼
 ✓ Preparing nodes 📦 📦 📦
 ✓ Writing configuration 📜
 ✓ Starting control-plane 🕹️
 ✓ Installing StorageClass 💾
 ✓ Joining worker nodes 🚜
Set kubectl context to "kind-antrea"
You can now use your cluster with:

kubectl cluster-info --context kind-antrea

Have a nice day! 👋
deployment.apps/coredns patched
Configuring networks
Using default kind docker network
load images
docker image projects.registry.vmware.com/antrea/antrea-ubuntu:latest not found

5. Observe that Antrea agent Pods never become ready

$ k get po -A
NAMESPACE            NAME                                           READY   STATUS                  RESTARTS       AGE
kube-system          antrea-agent-9hjqn                             0/2     Init:CrashLoopBackOff   5 (86s ago)    5m35s
kube-system          antrea-agent-w95zm                             0/2     Init:CrashLoopBackOff   5 (96s ago)    5m35s
kube-system          antrea-agent-wklfx                             0/2     Init:CrashLoopBackOff   5 (101s ago)   5m35s
kube-system          antrea-controller-7d8bbd857-9r6wx              1/1     Running                 0              5m36s
kube-system          coredns-6d4b75cb6d-b2f5j                       0/1     ContainerCreating       0              6m2s
kube-system          coredns-b4b5969d4-fqqdc                        0/1     ContainerCreating       0              5m41s
kube-system          coredns-b4b5969d4-lfnx5                        0/1     ContainerCreating       0              5m41s
kube-system          etcd-antrea-control-plane                      1/1     Running                 0              6m16s
kube-system          kube-apiserver-antrea-control-plane            1/1     Running                 0              6m16s
kube-system          kube-controller-manager-antrea-control-plane   1/1     Running                 0              6m19s
kube-system          kube-proxy-fpsxj                               1/1     Running                 0              6m3s
kube-system          kube-proxy-ttj26                               1/1     Running                 0              5m57s
kube-system          kube-proxy-vznxp                               1/1     Running                 0              5m44s
kube-system          kube-scheduler-antrea-control-plane            1/1     Running                 0              6m16s
local-path-storage   local-path-provisioner-9cd9bd544-7629z         0/1     ContainerCreating       0              6m2s

Expected

A KinD cluster is built with the specified version of Antrea and everything is in a healthy position. Actual behavior

Antrea does not become healthy and therefore cannot be fully used. Versions:

Please provide the following information:

• Antrea version (Docker image tag). v1.6.1
• Kubernetes version (use kubectl version). If your Kubernetes components have different versions, please provide the version for all of them. v1.24.0
• Container runtime: which runtime are you using (e.g. containerd, cri-o, docker) and which version are you using? Docker
• Linux kernel version on the Kubernetes Nodes (uname -r). 5.10.102.1-microsoft-standard-WSL2
• If you chose to compile the Open vSwitch kernel module manually instead of using the kernel module built into the Linux kernel, which version of the OVS kernel module are you using? Include the output of modinfo openvswitch for the Kubernetes Nodes.

Additional context

[chipzoller]

After reading the documentation further, once I manually pull the projects.registry.vmware.com/antrea/antrea-ubuntu:latest image and re-run the script, it complete successfully. There either needs to be:

1. A provision in the script to detect and pull (if missing) the appropriate image prior to attempting the load or;
2. a note in the documentation stating that is a prerequisite to running the kind-setup.sh script.

EDIT: I realize, as stated here, running Antrea on WSL2 backend isn't supported, so the point of this issue isn't making it work but fixing the scripted "quick start" approach to running Antrea.

[tnqn]

@chipzoller thanks for the details. kind-setup.sh is a convenience script typically used for testing a local build so it was made to not pull the image automatically. But as you suggested we could make it more friendly by detecting whether it's available locally before loading it and throwing error.

BTW, if loading local antrea image to kind nodes failed, the nodes would still work as they could pull the image from public repo. I guess anrea-agent went to Init:CrashLoopBackOff status because some required kernel modules in the WSL2 backend were missing as the doc https://github.com/antrea-io/antrea/blob/main/docs/kind.md#antrea-agents-are-not-starting-on-windows-what-could-it-be listed? If you could build a custom Kernel for WSL 2 with the listed modules, it should work.

[mohitsaxenaknoldus]

I can work on adding this functionality to the script. Please assign.

[tnqn]

thanks @mohitsaxenaknoldus, assigned it to you

[mohitsaxenaknoldus]

I can't pull the image:

knoldus@knoldus-Latitude-3510:~/OS/antrea/ci/kind$ sudo docker pull projects.registry.vmware.com/antrea/antrea-ubuntu:latest
latest: Pulling from antrea/antrea-ubuntu
d5fd17ec1767: Pull complete 
67ff7a99b9f4: Downloading 
41417343aff5: Download complete 
f891394d3432: Download complete 
fa8cd1e41c4f: Download complete 
e7e9f0ca0573: Download complete 
28f0b358a818: Download complete 
9cdeaef5e864: Download complete 
eec577fe2dad: Download complete 
unknown blob

[github-actions[bot]]

This issue is stale because it has been open 90 days with no activity. Remove stale label or comment, or this will be closed in 90 days