[PVE-User] Error deleting Ceph volume from KVM virtual machine

Greg Poirier greg.poirier at opower.com
Sun May 18 21:24:31 CEST 2014


Right. I understand, but not all of my Proxmox users have read
documentation. This was more of an experiment to see exactly what happens
when you try to do something like... remove a disk from a running VM.

The inconsistency of PVE's behavior is what I find most concerning. I will
continue to experiment and read through the source. If I can make some
improvements I will, but I stopped writing Perl a while back (by choice ;p).


On Sun, May 18, 2014 at 12:15 PM, Eneko Lacunza <elacunza at binovo.es> wrote:

>  Hi Greg,
>
> You should always stop a VM to effectively remove a hard disk: the running
> VM doesn't free it until it is stopped. You can make changes in configured
> harddisks in PVE GUI but they won't be activated until next stop/start
> (reboot doesn't work AFAIK).
>
> That said, the remove error could be because of a timeout deleting the
> somewhat big disk.
>
> Cheers
> Eneko
>
>
> On 17/05/14 00:44, Greg Poirier wrote:
>
> I have a 1TB RBD volume attached to a KVM virtual machine. Removing the
> volume from this machine was somewhat problematic.
>
>  While the machine is running:
>
>  I selected the disk and clicked Remove. This cause the disk to appear as
> 'unused0' which mapped to the Ceph RBD volume.
>
>  I then highlighted 'unused0' and clicked Remove again.
>
>  I got three errors:
> One was a timeout error attempting to get flock the VM's lock file
> (/var/lock/qemu-server/lock-400.conf).
> Another was a connection timeout error.
> Another was an internal server error.
>
>  I was unable to find any logs specifying the exact nature of any of
> these errors. pvedaemon and pveproxy both simply logged the request.
>
>  I tried deleting the disk from Ceph directly which is when I discovered
> the watcher on the rbd_header object. I verified that it was the KVM
> process that held the lock on the RBD volume and shutdown the VM.
>
>  While the machine is not running:
>
>  After selecting the 'unused0' disk and clicking Remove the web interface
> froze for some time, and then finally came back with an Internal Server
> Error.
>
>  The volume was no longer present in the web interface, and I confirmed
> both that the volume was no longer listed in rbd -p volumes ls -- as well
> as inspecting attempting to inspect all of the objects that used to make up
> the volume (rbd.stat returned an exception saying that the file wasn't
> found--confirming the volume's completed deletion).
>
>  I am running PVE 3.1-24/060bd5a6  with the 2.6.32 kernel.
>
>  Is this a known issue? Has this been fixed in later versions of Proxmox?
> Should I file a new bug? Is there any additional information I should
> attempt to gather before filing?
>
>  Thanks!
>
>
> _______________________________________________
> pve-user mailing listpve-user at pve.proxmox.comhttp://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user
>
>
>
> --
> Zuzendari Teknikoa / Director T├ęcnico
> Binovo IT Human Project, S.L.
> Telf. 943575997
>       943493611
> Astigarraga bidea 2, planta 6 dcha., ofi. 3-2; 20180 Oiartzun (Gipuzkoa)www.binovo.es
>
>
> _______________________________________________
> pve-user mailing list
> pve-user at pve.proxmox.com
> http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://pve.proxmox.com/pipermail/pve-user/attachments/20140518/05f21f77/attachment-0015.html>


More information about the pve-user mailing list