Re: cluster-glue_1.0.9+hg2665-1~bpo60+2_amd64: lrmd deadlocks
Am 14.03.12 04:02, schrieb Phillip Frost:
> Trying to use pacemaker and dependencies from squeeze-backports, there seems to be a known interaction with glib and lrmd that results in some deadlocks. After starting corosync and pacemaker, just about anything I ran would deadlock, including adding resources, and even running "crm ra classes" or stopping pacemakerd. See here:
>
> http://oss.clusterlabs.org/pipermail/pacemaker/2012-March/013439.html
>
> this seems to be a known bug:
>
> https://bugs.launchpad.net/ubuntu/oneiric/+source/cluster-glue/+bug/821732
>
> I was able to make a build that leaks sockets instead of deadlocking by applying debian/patches/raexecupstart.patch, which seems to be disabled in squeeze-backports. According to the changelog, it seems it was disabled because the underlying bug in glib was fixed in unstable, but I'm guessing it's not fixed in squeeze.
>
> Would be great if it didn't deadlock *and* didn't leak sockets. However, I wasn't able to sort out the conversation on launchpad enough to accomplish that, or even determine if it's possible.
>
Hello there,
can you please double-check whether the problem really still exists with
cluster-glue_1.0.9+hg2665-1~bpo60+2? The only change between +1 and this
version was to disable upstart altogether; upstart support is just not
present in +2. If it still deadlocks because of that glib bug, then we
definetely have some other problem to watch out for.
Best regards
Martin
--
Martin Gerhard Loschwitz
Debian GNU/Linux Developer
Reply to: