On 27/08/2013, at 3:46 PM, Nate Clark <[email protected]> wrote: > > From: Andrew Beekhof <[email protected]> > To: The Pacemaker cluster resource manager <[email protected]> > Cc: > Date: Mon, 26 Aug 2013 09:26:32 +1000 > Subject: Re: [Pacemaker] Pacemaker 1.1.7 and negative corosync nodeid > > On 24/08/2013, at 5:26 AM, Nate Clark <[email protected]> wrote: > > > I have a few systems running pacemaker 1.1.7 using corosync and some of > > them hit the issue that the nodeid in the cib is recorded as a negative > > number. This issue was resolved in commit > > f0e9b262f1d3219df248a5340771f8e23c5d6bce. However if you upgrade pacemaker > > to 1.1.8, the new pacemaker now sees two nodes for every node that hit this > > issue. It sees one with the negative nodeid and one with the positive. > > > > Is there a way to prevent this from happening when upgrading from 1.1.7? > > One way I can think of doing this would be to edit the cib while pacemaker > > is down so that when it comes up the values are now positive but I am not > > sure of a way to do that because of the cib signing. > > Once the cluster is running again, you can "just" clear out the old entries > with: cibadmin --replace --xml-text '<nodes/>' > That should be enough IIRC > > > That seems to work fine if you upgrade all nodes simultaneously but if you > try and perform a rolling upgrade the negative node ids come back. I assume > this is because the node running the old pacemaker is still using those > nodeids and is sending them back to the new node. Is there anyway to be able > to perform this upgrade rolling or will it have to be all nodes at once?
All nodes at once I think. You can run the cibadmin command without a cluster running though, perhaps that will help. You'll need to prepend: CIB_file=/path/to/cib.xml
signature.asc
Description: Message signed with OpenPGP using GPGMail
_______________________________________________ Pacemaker mailing list: [email protected] http://oss.clusterlabs.org/mailman/listinfo/pacemaker Project Home: http://www.clusterlabs.org Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf Bugs: http://bugs.clusterlabs.org
