2.0.5 is quite old and has quite a few bugs!

I suggest you try out the newest release (2.1.2) first and see if that somewhat helps you. If that doesn't help you could ask again here including your config files and a detailed description... :)


greets Lars.


Shailesh schrieb:
Hi,
     I have not recieved any responses to my issue yet, I am of the
feeling  that the solution to this problem is straight-forward and that
some experts on the list can giude me out of this. Hope you can oblige.

Thanks & Regards
Shailesh

On Tue, 2007-07-31 at 12:05 +0530, Shailesh wrote:
Hello All,
I would appreciate if you could help me on this problem I am facing with Apache HA with HB and MON.

        I have been working on setting up 2 node failover cluster for my
web service. I have installed the heartbeat 2.0.5 amd MON on the 2 SUSE
Linux servers. The MON is monitoring the Apache webserver. I tested two
methods  of causing failover and then a failback. I end up having a
split brain in the cluster in Method 1.


Method 1:

I find that SLAVENODE takes all the resource if I stop the heartbeat of
the MASTERNODE by running 'rcheartbeat stop', this is quite normal.
But If I do 'rcheartbeat start' on the MASTERNODE again to restart
heartbeat, the MASTERNODE thinks the SLAVENODE is dead and takes over
the resources ending up in a unrecoverable split-brain.
Method 2:
Suprisingly, If I had caused the failover by pulling off the network
cable and the restored back the network cable followed by starting the
heartbeat again on the MASTERNODE,  I see that MASTERNODE senses the
SLAVENODE, SLAVENODE relinquishes resources to MASTER and it seems all fine.

I am not able to get why the Method-1 of failover is ending up with
a split brain.

My ha.cf and haresource are as below.
debug 1
logfile /var/log/ha-log
keepalive 2
warntime 30
deadtime 80
initdead 90
node MASTERNODE
node SLAVENODE
bcast eth0
udpport 694
auto_failback on
ping_group ping-cluster-test 10.10.10.1 10.10.10.151
respawn hacluster /usr/lib/heartbeat/ipfail
crm off

Also attached are the master and slave dump when split brain occurs in
Method-1.

It would be great to get your solutios to this.


Regards
Shailesh P Shirali







_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to