Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
We have a large number of xen servers with virtual hosts running
mostly Debian lenny, with 2.6.18.8 kernels from xensource. Recently
we've been experimenting with pacemaker, as a result of some split
brain problems caused on some of our old heartbeat drbd servers.
Since this problem can be reproduced using drbd alone, without
involving pacemaker, I'll just concentrate on that part of the
software:
Kernel: linux 2.6.18.8-xen version 3.4.0, from xensource
also with an older version of 2.6.18.8-xen from xen
Xen hypervisor: version 3.2.1, from Debian sources
drbd: version 8.3.4-12 from git.drbd.org
also with version 8.3.4-1 from Debian sources
64-bit architecture
The symptom: installed drbd, edited configure file, and ran
drbdadm create-md r0
drbdadm attach r0
drbdadm syncer r0
drbdadm connect r0
on each machine. No problems. Then ran
drbdadm -- --overwrite-data-of-peer primary r0
This started to sync the 20-gb disks, and continued until about 5% of
the disk was synced. Then the xen dom0 rebooted. The problem is
reproduceable.
I see there have been other problems reported of kernel panics during
syncronisation. Is this likely to be the same problem? Is there a
fix?
Owen Le Blanc
University of Manchester