I can open a bug on this. Should I file it for the entire "snapshot removal fails"-problem or just this error? I've looked at the parsing error and the problem seems to be that the qemuImg.check() function expects a qemu-img version that outputs a string like: "Image end offset: 262144". The only version I could find that does that is the one in the F19 virt-preview repository. Stock F19 qemu-img doesn't have this output. Nor do the stock EL6 or qemu-img-rhev EL6 versions. The only function that calls qemuImg.check is blockVolume.shrinkToOptimalSize() this function needs (and uses) the parsed offset to shrink the LV containing the qcow2 image. On Fri, Nov 8, 2013 at 4:33 PM, Federico Simoncelli <fsimonce@redhat.com> wrote:
----- Original Message -----
From: "Dan Kenigsberg" <danken@redhat.com> To: "Sander Grendelman" <sander@grendelman.com> Cc: users@ovirt.org, fsimonce@redhat.com, ykaplan@redhat.com Sent: Friday, November 8, 2013 4:06:53 PM Subject: Re: [Users] Live storage migration snapshot removal (fails)
On Fri, Nov 08, 2013 at 02:20:39PM +0100, Sander Grendelman wrote:
<snip>
9d4e8a43-4851-42ff-a684-f3d802527cf7/c512267d-ebba-4907-a782-fec9b6c95116 52178458-1764-4317-b85b-71843054aae9::WARNING::2013-11-08 14:02:53,772::image::1164::Storage.Image::(merge) Auto shrink after merge failed Traceback (most recent call last): File "/usr/share/vdsm/storage/image.py", line 1162, in merge srcVol.shrinkToOptimalSize() File "/usr/share/vdsm/storage/blockVolume.py", line 320, in shrinkToOptimalSize qemuImg.FORMAT.QCOW2) File "/usr/lib64/python2.6/site-packages/vdsm/qemuImg.py", line 109, in check raise QImgError(rc, out, err, "unable to parse qemu-img check output") QImgError: ecode=0, stdout=['No errors were found on the image.'], stderr=[], message=unable to parse qemu-img check output
I'm not sure that it's the only problem in this flow, but there's a clear bug in lib/vdsm/qemuImg.py's check() function: it fails to parse the output of qemu-img.
Would you open a bug on that? I found no open one.
I remember that this was discussed and the agreement was that if the offset is not reported by qemu-img we should have used the old method to calculate the new volume size.
We'll probably need to verify it. Sander can you open a bug on this?
Thanks, -- Federico