Shuo Wu

Do you mean the XFS filesystem is corrupted when the issue is triggered? Or is there inconsistency among replicas of the volume?

> There is inconsistency among replicas. In the two times I've seen it, one replica differs in its hash for a given volume. Following the guide to delete the inconsistent...

Weird... What's the underlying filesystem of the replica directories? We may need to try reproducing it first.

Can you export the kubelet log and `dmesg` result besides the Longhorn support bundle? I am not sure if there are more clues in these logs but let's take a...

@jperville I guess the cause of this issue is that the VM & Longhorn volume is shut down while the metadata of the underlying fs is being modified. Then this...

> At this point it's becoming an undue annoyance, and I'm going to copy everything remaining out, re-init all Longhorn volumes as ext4, copy back, and move on. Sure, you...

> There was no pod accessing the volume for a period of days OK. Now the main issue in your case is, why the volume suddenly becomes inaccessible. Is there...

Right now the recurring job is separated from the volume as a new CRD. Hence backing up recurring jobs involves re-considering the backup data/file format. This would take some time.

How many snapshots does the volume have? Actual size = the sum of the size of all snapshot files Parent size/The size you set when you create a volume =...

I guess what you did is writing then deleting files in the filesystem of the volume, and the created files, in the end, are few Mi. If this is the...