E2E failures post cluster-api version bump to v1.10

[Amulyam24]

• Cluster cleanup failing with - Add kubeconfig and cluster config path while deleting the cluster in e2e #2308

{input.KubeConfigPath is required for DumpAllResources
Expected
    <string>: 
not to be empty failed [FAILED] input.KubeConfigPath is required for DumpAllResources
Expected
    <string>: 
not to be empty
In [AfterEach] at: /home/prow/go/pkg/mod/sigs.k8s.io/cluster-api/[email protected]/framework/alltypes_helpers.go:132 @ 04/11/25 00:24:32.144
}

• PowerVS cluster creation failing while applying cluster template - Update cluster-api to v1.10.1 #2334
  FIx has been added to cluster-api v1.10.1

 Failed to apply the cluster template
  Expected success, but got an error:
      <errors.aggregate | len:2, cap:2>:
      [KubeadmConfigTemplate.bootstrap.cluster.x-k8s.io "capibm-e2e-2nmy7v-md-0" is invalid: spec.template.spec.joinConfiguration.discovery.bootstrapToken.token: Invalid value: "": spec.template.spec.joinConfiguration.discovery.bootstrapToken.token in body should be at least 1 chars long, KubeadmControlPlane.controlplane.cluster.x-k8s.io "capibm-e2e-2nmy7v-control-plane" is invalid: spec.kubeadmConfigSpec.joinConfiguration.discovery.bootstrapToken.token: Invalid value: "": spec.kubeadmConfigSpec.joinConfiguration.discovery.bootstrapToken.token in body should be at least 1 chars long]

• VPC cluster creation fails with load balancer stuck in update_pending state - infra issue and not because of version bump, will be handled separately

E0416 00:26:16.467053       1 controller.go:347] "Reconciler error" err="failed to bind port 6443 to control plane create-workload-cluster-u80301/capibm-e2e-62ojyw-control-plane-7jstc: error load balancer is not in active state"

• VPC cluster creation fails during subnet creation - Change the number of ginkgo nodes to 1 to run the e2e tests serially #2323

"Reconciler error" err="failed to reconcile Subnet for IBMVPCCluster create-workload-cluster-v0k383/capibm-e2e-mgb75f: Subnet CIDR 10.250.0.0/18 conflicts with existing subnet: 02t7-2bdd8e00-4542-41a3-ae73-b37a52f4fbe8" controller="ibmvpccluster" controllerGroup="infrastructure.cluster.x-k8s.io" controllerKind="IBMVPCCluster" IBMVPCCluster="create-workload-cluster-v0k383/capibm-e2e-mgb75f" namespace="create-workload-cluster-v0k383" name="capibm-e2e-mgb75f" reconcileID="1e481968-17cb-4dba-830b-c8a45cfc68c1"

• Log collector is nil - minor,not blocking the flow

Captured StdOut/StdErr Output >>
  Unable to get logs for workload Cluster create-workload-cluster-o1y9sj/capibm-e2e-1axqky: log collector is nil.
  << Captured StdOut/StdErr Output

[Amulyam24]

Regarding VPC cluster creation with subnet failure, it is occuring as 2 cluster creations are being triggered simultaneously in the same region during the e2e run causing the conflict

NAMESPACE                        NAME                CLUSTERCLASS   PHASE          AGE    VERSION
create-workload-cluster-kipc25   capibm-e2e-czpvdc                  Provisioning   2m1s
create-workload-cluster-v0k383   capibm-e2e-mgb75f                  Provisioning   2m

Need to identify why is it happening.

[mkumatag]

2 cluster creations are being triggered simultaneously in the same region

how is this even possible? boskos is suppose to allocate a different region right?

[Amulyam24]

2 cluster creations are being triggered simultaneously in the same region

how is this even possible? boskos is suppose to allocate a different region right?

In e2e, one region is acquired per job. We have test cases for 2 cluster creations which usually run serially. As per the logs, I suspect they are being triggered together and hence causing the issue.

[Amulyam24]

Changing the nodes to 1 fixes the issue but not sure how we haven't encountered this until now when the flag was being set since the start.

[Amulyam24]

/close

[k8s-ci-robot]

@Amulyam24: Closing this issue.

In response to this:

/close

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository.