Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Lars Ellenberg wrote: > / 2006-04-28 15:07:48 -0500 > \ Christopher Harrison: > >> I love drbd and the ease of use. Kudos. >> Lately I have been having problems with the drbd block device. The >> problem manifests in closing a unix socket on the drbd device device. >> The result freezes all io to the drbd devices (both primary and >> secondary). The drbd device lock appears unrecoverable. I have >> found a way to re sync the systems but it requires a reboot of both >> machines to re sync the devices. After the reboot I force the >> secondary to be primary and blow away the meta data from the former >> primary device. Only after I destroy the meta data can I start my >> re sync to the former primary. If I try to re sync with the old meta >> data, the previous primary system locks up again when it tries to >> become primary again. >> > > I seriously doubt that this has anything to do with unix domain sockets. > Anything else that may be the cause of your troubles? > What exactly do you do to reproduce the problem? > In case you have a simple step-by-step starting "from scratch", > that leads you intro trouble always, we probably can reproduce this on > some testcluster here, which would really help in debugging. > > To reproduce the problem I installed postfix (simple install with FC5 and yum) then I copied the /var/lib/postfix directory to the drbd partition changed the postfix config to point the configdirectory and partition-default to the drbd partition. Then I started postfix. Postfix started just fine and there did not appear to be any problems on startup but when I tried to kill postfix (stop and kill) the system just hung. I was still able to login and start another ssh session but the system load increased and the situation ensured. I hope this helps. -Chris