Related discussion
https://github.com/longhorn/longhorn/issues/3907
Test CSI plugin liveness probe should recover CSI socket file
Given healthy Longhorn cluster
When delete the Longhorn CSI socket file on one of the node(node-1
).
rm /var/lib/kubelet/plugins/driver.longhorn.io/csi.sock
Then the longhorn-csi-plugin-*
pod on node-1
should be restarted.
And the csi-provisioner-*
pod on node-1
should be restarted.
And the csi-resizer-*
pod on node-1
should be restarted.
And the csi-snapshotter-*
pod on node-1
should be restarted.
And the csi-attacher-*
pod on node-1
should be restarted.
Note: In
node-1
, the other pods of the CSI sidecar components restarted are triggered by their internal ProbeForever to the CSI connection. It is not due to the liveness probe of thelonghorn-csi-plugin
.
And Longhorn CSI socket file should exist.