On 07/01/10 21:58, Evan Broder wrote: > On Mon, Jan 4, 2010 at 12:37 AM, Quentin Smith<[email protected]> wrote: >> [please preserve the CC when replying, thanks] >> >> Hi- >> >> I'm trying to set up a new cluster with OpenAIS 1.1.1, Corosync 1.1.2, and >> LVM 2.02.56. Whenever I try to bring up the cluster (using aisexec, which >> sets the COROSYNC_DEFAULT_CONFIG_IFACE environment variable and then runs >> corosync), the nodes all seem to immediately find each other based on the >> debugging output, but no other programs are able to use the cluster. In >> particular, clvmd startup fails with: >> >> r...@black-mesa:/etc/corosync# clvmd -d >> CLVMD[6f0d780]: Jan 4 01:33:01 CLVMD started >> CLVMD[6f0d780]: Jan 4 01:33:01 Cannot join clvmd process group: 6 >> CLVMD[6f0d780]: Jan 4 01:33:01 Can't initialise cluster interface >> Can't initialise cluster interface >> >> and even corosync-cfgtool fails with the same error: >> >> r...@black-mesa:/etc/corosync# corosync-cfgtool -s >> Printing ring status. >> Local node ID 16779018 >> Could not get the ring status, the error is: 6 >> >> Looking at the source, I see that error 6 is "SA_AIS_ERR_TRY_AGAIN". I'm not >> sure how long I'm expected to wait for the cluster to start functioning; >> I've already tried waiting 5 minutes, which seems like it should be long >> enough. >> >> I've also tried configuring pacemaker (by adding a service block to >> /etc/corosync/corosync.conf), and it starts functioning within 30 seconds of >> bringing up corosync, even while clvmd and corosync-cfgtool refuse to start. >> >> When I don't set the COROSYNC_DEFAULT_CONFIG_IFACE environment variable and >> just load the openais_lck service, the cluster does eventually start >> working, and I can even start clvmd, but the locking appears to be >> nonfunctional - running two lvcreate operations simultaneously causes a >> deadlock every time. >> >> I've attached a copy of my configuration file below. Is there something >> missing in my configuration, or is this a bug in OpenAIS/Corosync? >> >> (The reason we're trying to get CLVM + OpenAIS working at all is that we've >> experienced numerous deadlocks with clvmd on dlm, which we are already >> running.) >> >> Thanks, >> --Quentin >> >> r...@black-mesa:/etc/corosync# cat corosync.conf >> compatibility: none >> >> totem { >> version: 2 >> secauth: off >> interface { >> ringnumber: 0 >> bindnetaddr: 10.7.0.0 >> mcastaddr: 226.94.1.1 >> mcastport: 5405 >> } >> } >> >> logging { >> to_syslog: yes >> syslog_facility: daemon >> debug: on >> timestamp: on >> } >> >> amf { >> mode: disabled >> } >> service { >> name: pacemaker >> ver: 0 >> } >> >> > > As a followup on this, we were able to get everything working using > aisexec after we set compatibility: whitetank in the configuration > file. > > We're still having problems with deadlocking when we try to do LVM > manipulations on both nodes simultaneously, but I suspect that's a > clvmd problem, not an openais one.
The whole point of clvmd is that it serialize LVM operations on nodes across the cluster. If you have more information about this can you post it to [email protected] please ? Chrissie _______________________________________________ Openais mailing list [email protected] https://lists.linux-foundation.org/mailman/listinfo/openais
