Environment:
Kubespray version: 2.14.1
Anything else do we need to know:
The cluster.yml playbook executed successfully, but pods from ebs-csi-controller deployment were in CrashLoopBackOff status. The problem is with csi-snapshotter container:
# kubectl logs -n kube-system ebs-csi-controller-bd7fdfdd8-dqp84 csi-snapshotter
flag provided but not defined: -connection-timeout
Usage of /csi-snapshotter:
...
After some investigation I have found out that the --connection-timeout flag has been deprecated in release 1.1.0 of kubernetes-csi/external-snapshotter and removed in release 2.0.0. The version of csi-snapshotter was updated from 1.2.0 to 2.1.1 in Kubespray 2.14.0. The flag does not exist in the new csi-snapshotter but remained in the Kubespray yamls https://github.com/kubernetes-sigs/kubespray/blob/v2.14.1/roles/kubernetes-apps/csi_driver/aws_ebs/templates/aws-ebs-csi-controllerservice.yml.j2#L92.
@rafal-jan thank you for the issue, care to make a PR ? :)
@floryut sure, but right now I am experiencing issues related to upgrade of csi-attacher (similar to https://github.com/kubernetes-sigs/kubespray/pull/6358#issuecomment-652448971). I will create a PR once the CSI driver is fully working.
Most helpful comment
@floryut sure, but right now I am experiencing issues related to upgrade of
csi-attacher(similar to https://github.com/kubernetes-sigs/kubespray/pull/6358#issuecomment-652448971). I will create a PR once the CSI driver is fully working.