kubernetes / kops

Kubernetes Operations (kOps) - Production Grade k8s Installation, Upgrades and Management
https://kops.sigs.k8s.io/
Apache License 2.0
15.91k stars 4.65k forks source link

Clusters created with Kubernetes 1.14 don't start #9604

Closed hakman closed 4 years ago

hakman commented 4 years ago

After latest bulk of merges, clusters fail to start for Kubernetes 1.11 - 1.14 and periodic tests fail: https://testgrid.k8s.io/kops-versions#kops-aws-k8s-1.14

/cc @rifelpet @johngmyers

rifelpet commented 4 years ago

from the job output:

2020/07/20 14:44:17 process.go:153: Running: kubectl get nodes -ojson
The connection to the server localhost:8080 was refused - did you specify the right host or port?

That looks like the kubeconfig file is not being set properly. Is there a reason https://github.com/kubernetes/test-infra/pull/18373 was closed?

rifelpet commented 4 years ago

/kind bug

hakman commented 4 years ago

@rifelpet "admin" user is added automatically when "create cluster --yes" is used, that's why I closed it.

johngmyers commented 4 years ago

But the tests are doing create cluster without --yes. In that case no kubecfg is created. So kubernetes/test-infra#18373 is the correct fix.

hakman commented 4 years ago

@johngmyers you are partially correct. Tests for 1.15+ are run with a newer version of kubetest that creates a cluster with --yes. Older tests seem to run an earlier version that runs update cluster --yes. Maybe we should just move all tests to latest kubekins-e2e image?