Red Hat Bugzilla – Bug 1254925
Snapshot merge leaves a forever running task
Last modified: 2016-03-10 01:15:05 EST
Created attachment 1064692 [details]
Logs from a setup exhibiting the problem
Description of problem:
Removing a snapshot of a VM succeeds, but after everything is done, there is still a task in the task list
Version-Release number of selected component (if applicable):
nightly master engine, nightly master vdsm, from 18.08.15
Steps to Reproduce:
1. Create VM from template (on NFS)
2. Start VM
3. Create 2 snapshots
4. Delete the first one
Snapshot is deleted, all disks active, task running forever
No task after snapshot is done
Greg, this looks awfully familiar... Can you take a look please?
I had the same problem in the same version, but with iSCSI storage and a VM that I created from scratch, with only one snapshot. Looking at the libvirt xml with virsh -r revealed that the snapshot was created correctly. When deleting, a block-pull job would run properly and the snapshot would merge. However, after the merge was complete in the libvirt xml, the snapshot disk would be left behind in /dev/<uuid>/, and the task would appear to run forever. In one case, the disk space was freed up, and the snapshot data removed, but the task still ran forever.
This does indeed appear to be the same issue that happened in 3.5. I tracked down the files and the code that was added in https://github.com/oVirt/ovirt-engine/commit/209ec823a03dd5838eed3d711fd821d2a1aba9dd is missing from 3.6. The symptoms are the same as bug https://bugzilla.redhat.com/1127464, which is now impacting 3.6.
My previous comment was incorrect. The code referenced was a deletion, and this carries through. Please disregard.
Created attachment 1067279 [details]
I've uploaded a copy of the engine log where this is happening. In this case, a cleanly rebuilt the cluster. I'm using iSCSI storage.
I was able to take snapshots and delete them with the VM off successfully.
I was also able to live move a disk from one domain to another.
The task only had problems when I tried deleting a snapshot with the VM running. In this case, the snapshot was actually deleted, but the interface task ran on. I will attached output from virsh -r showing the libvirt xml before and after, as well as the log.
Created attachment 1067280 [details]
output of virsh -r, corresponding to the above engine log.
Note. In the most recent case, rebooting the ovirt-engine server cleared the message in the interface. After having this clear once, I was able to create and delete a snapshot successfully with the VM running.
Based on comment #7