This bug is a really serious issue for us. Let me explain. We run active/passive MySQL-clusters using Pacemaker and DRBD. For backups, we create an LVM-snapshot on the passive node, fire up a MySQL server and dump data from that snapshot. When the dump is finished, we stop MySQL on the passive node and remove the snapshot. This fails almost every time due to this bug.
The big problem though is that when the bug appears and the LVM subsystem freezes, our active node hangs completetly in iowait, due to DRBD unable to finish IO operations on the passive node. So basically the whole cluster fails and we get hard downtime as long as we don't take action manually and do 'lvresume' and/or reboot. We had a look on some of the proposed workarounds, but they are no option for us. What i basically want to ask is - are there any ETAs possible when this will be resolved? We are seriously considering a downgrade to squeeze. That would be a _huge_ pain, though. Thanks, Frank
