Re: corosync 1.4.2 membership problem

[Andrew Beekhof <andrew@xxxxxxxxxxx>]

On Wed, Feb 29, 2012 at 4:35 PM, Kiss Bence <bence@xxxxxxxxxxx> wrote:
> Hi,
>
> I have run into some trouble with corosync on Fedora 14 systems.
> I'm trying to build a cluster. First I thougth it is a pacemaker bug so I
> wrote there and also submitted a bugreport, but it turns to be out maybe a
> corosync bug.

Essentially corosync sits there looping like this for about 5 minutes:

Feb  1 19:30:00 virt1 corosync[943]:   [TOTEM ] A processor joined or
left the membership and a new membership was formed.
Feb  1 19:30:00 virt1 corosync[943]:   [CPG   ] chosen downlist:
sender r(0) ip(192.168.122.2) ; members(old:1 left:0)
Feb  1 19:30:00 virt1 corosync[943]:   [MAIN  ] Completed service
synchronization, ready to provide service.
Feb  1 19:30:01 virt1 corosync[943]:   [TOTEM ] A processor joined or
left the membership and a new membership was formed.
Feb  1 19:30:01 virt1 corosync[943]:   [CPG   ] chosen downlist:
sender r(0) ip(192.168.122.2) ; members(old:1 left:0)
Feb  1 19:30:01 virt1 corosync[943]:   [MAIN  ] Completed service
synchronization, ready to provide service.
Feb  1 19:30:03 virt1 corosync[943]:   [TOTEM ] A processor joined or
left the membership and a new membership was formed.
Feb  1 19:30:03 virt1 corosync[943]:   [CPG   ] chosen downlist:
sender r(0) ip(192.168.122.2) ; members(old:1 left:0)
Feb  1 19:30:03 virt1 corosync[943]:   [MAIN  ] Completed service
synchronization, ready to provide service.



> The cluster in real env and also in test env (virtual machines on the same
> host) shows that after killing forcibly one of the node, the communication
> between the nodes is restored only after a while (about 5 minutes).
> Meanwhile the ip, the tcp and the udp connections are well. The drbd
> subsystem is intact. The logs shows that both nodes does not see the other
> in that particular period of time.
>
> The whole discussion can be read here:
>
> http://www.mail-archive.com/pacemaker@xxxxxxxxxxxxxxxxxxx/msg11278.html
> http://bugs.clusterlabs.org/show_bug.cgi?id=5031
>
> [root@virt1 ~]# rpm -qi corosync
> Name        : corosync                     Relocations: (not relocatable)
> Version     : 1.4.2                             Vendor: Fedora Project
> Release     : 1.fc14                        Build Date: Thu 06 Oct 2011
> 10:37:19 AM CEST
> Install Date: Fri 28 Oct 2011 04:12:14 PM CEST      Build Host:
> x86-06.phx2.fedoraproject.org
> Group       : System Environment/Base       Source RPM:
> corosync-1.4.2-1.fc14.src.rpm
> Size        : 414020                           License: BSD
> Signature   : RSA/SHA256, Thu 06 Oct 2011 11:48:08 AM CEST, Key ID
> 421caddb97a1071f
> Packager    : Fedora Project
> URL         : http://ftp.corosync.org
> Summary     : The Corosync Cluster Engine and Application Programming
> Interfaces
>
> Thanks,
> Bence
>
> _______________________________________________
> discuss mailing list
> discuss@xxxxxxxxxxxx
> http://lists.corosync.org/mailman/listinfo/discuss
_______________________________________________
discuss mailing list
discuss@xxxxxxxxxxxx
http://lists.corosync.org/mailman/listinfo/discuss