On 02:43 Mon 04 Nov , [email protected] wrote: > Hello all, > > it seems that we are facing the same problem on our ganeti clusters that > have been upgraded from squeeze to wheezy. The live migration failure ratio > is huge, 1 out of 8-10 migrations fail. When these nodes where running > squeeze there were absolutely no such issues. > For comparison purposes here are the package details of our clusters: > > Squeeze: > drbd8-utils: > Installed: 2:8.3.7-2.1 > ganeti2: > Installed: 2.4.2+ippool5-1 > linux-image-2.6.32-5-amd64: > Installed: 2.6.32-48squeeze3 > > Wheezy: > drbd8-utils: > Installed: 2:8.3.13-2 > ganeti2: > Installed: 2.8.1-1~bpo70+httpboot > linux-image-3.2.0-4-amd64: > Installed: 3.2.51-1 > > I am attaching you the relevant files of two failed migrations. To > replicate the issue, just migrate a VM in while true; loop...and wait for a > little bit. > We have experienced the same issue with ganeti 2.6, 2.7 and 2.8, but also > with squeeze and kernel 3.2 from backports. > The common denominator in all problematic situations seems to be kernel 3.2 > but maybe there's a way to overcome this issue in ganeti itself. > > Can someone with better insight of drbd/ganeti/kernel take a look at the > proposed "option a" fix from: > http://lists.linbit.com/pipermail/drbd-user/2013-July/020173.html would > that work?
Ok, it looks like we got it! Basically the thread above is correct, an analysis and patch will follow. Regards, Apollon
