When working with 150+ volumes publishing / unpublishing at the same time, the current timeout logic causes grpc timeout errors to be seen by the csi-attacher process
This PR updates the timeout down to 5 seconds (from effectively 100s in waitForVolumeStatus) and hands off retry logic to the csi-attacher process.
In addition, the fixes a few logic bugs around unpublish when NodeID is passed as an arg in NodeUnpublishVolumeRequest and set to a different node to the currently attached node. This should safely clear out dangling VolumeAttachments.
When working with 150+ volumes publishing / unpublishing at the same time, the current timeout logic causes grpc timeout errors to be seen by the csi-attacher process
This PR updates the timeout down to 5 seconds (from effectively 100s in waitForVolumeStatus) and hands off retry logic to the csi-attacher process.
In addition, the fixes a few logic bugs around unpublish when NodeID is passed as an arg in
NodeUnpublishVolumeRequest
and set to a different node to the currently attached node. This should safely clear out dangling VolumeAttachments.