pod/spiderpool-agent "CrashLoopBackOff" error at the edge nodes when trying to install edgemesh using helm with enableCNI enabled

[qpanpony]

Trying to install edgemesh using helm with custom values.yaml (with enableCNI enabled).

pod/edgemesh sucessfully runs in both cloud and edge nodes, however pod/spiderpool-agent at the edgeside nodes show "CrashLoopBackOff" error.

I've checked edgecore logs at one of the edge nodes ().

Seems some CNI-releated issues at the edge nodes.
Any Suggestions? Thank you.

[Poorunga]

edgemesh-CNI is still in testing, you can disable it in helm.

[Poorunga]

/cc @NKDYX

[shyer]

我也遇到了这个问题，这是边缘POD无法访问集群API导致，我根据以下文档进行了相关配置。
https://edgemesh.netlify.app/zh/guide/edge-kube-api.html#%E8%83%8C%E6%99%AF
https://kubeedge.io/zh/docs/advanced/inclusterconfig/

可以访问集群API了，但是安装还是失败。调试发现，需要请求https://169.254.30.10:10550/readyz?timeout=2s

请求如何配置权限。这里的rabc也添加了。
https://github.com/kubeedge/kubeedge/blob/master/manifests/charts/cloudcore/templates/rbac_cloudcore_feature.yaml

[Icarus9913]

https://github.com/spidernet-io/spiderpool/blob/335952e3e920d9df0ec943489386b676d29d6bb7/cmd/spiderpool-agent/cmd/daemon.go#L287-L326

spiderpool-agent在启动的时候会去探测以下api-server的readyz路由，看起来似乎是访问到了clusterIP ？
@shyer

[shyer]

我调试，它访问的是：https://169.254.30.10:10550/readyz?timeout=2s，报401。
然后我手动请求clusterIP，curl -k https://10.96.0.1:443/readyz，反倒是正常的。

在kubeedge也提了一个issue：
kubeedge/kubeedge#5961

[Icarus9913]

这个169网段是pod IP还是？根据上面那个代码，我怀疑是 config := ctrl.GetConfigOrDie()这个 ctrl "sigs.k8s.io/controller-runtime" SDK 读的环境变量还是哪里拿的的这个IP地址, 应该是个配置问题

[shyer]

169.254.30.10 是kubeedge metaserver的地址，问题应该是通过metaserver请求readyz异常导致的（有提交了issue）

https://github.com/kubeedge/kubeedge/blob/master/docs/proposals/inclusterconfig.md#non-hostnetwork-edge-pods-access-metaserver