On Tue, 10 Aug 2010, fengyandong wrote: > On Mon, Aug 9, 2010 at 7:55 PM, Dejan Muhamedagic <[email protected]>wrote: > >> Hi, >> >> On Mon, Aug 09, 2010 at 09:27:14AM +0800, fengyandong wrote: >>> Thanks for your help. >>> >>> You means that if I had not stonith defined, 'restart network service ' >> will >>> must cause brainsplit in heaertbeat 2.1.4? >>> Could you tell me why 'restart network service ' causes brainsplit. >> >> Because that way you interrupt the communication between the nodes. >> > Yes, the communication between the nodes may be interrupt during the > network restart, but after the network recovers, why the communication > could not recover? I use *tcpdump* to find that the nodes could receive the > heartbeat package each other after the network recovers.
they do recover, but if the network outage is longer than your deadtime, the inactive box has decided the other one is dead and become active, so when the network is restored you are in a splitbrain situation, and that is what you are recovering from. David Lang >> >> Thanks, >> >> Dejan >> >>> Thanks. >>> >>> On Fri, Aug 6, 2010 at 8:29 PM, Dejan Muhamedagic <[email protected] >>> wrote: >>> >>>> Hi, >>>> >>>> On Fri, Aug 06, 2010 at 04:57:43PM +0800, fengyandong wrote: >>>>> Hi, >>>>> >>>>> I encounter a problem that 'restart network service ' casues >> brainsplit >>>> >>>> Not a good idea to restart network interfaces. If you had stonith >>>> defined, which you should, one of your nodes would've been >>>> fenced. >>>> >>>>> Heartbeart version 2.1.4 >>>> >>>> You need to upgrade to Pacemaker 1.0 or 1.1. >>>> >>>> Thanks, >>>> >>>> Dejan >>>> >>>>> The content of */etc/ha.d/ha.cf:* >>>>> initdead 20 >>>>> crm on >>>>> auto_failback off >>>>> deadping 10 >>>>> logfacility local0 >>>>> logfile /var/log/ha-log >>>>> deadtime 5000ms >>>>> keepalive 1000ms >>>>> ucast eth0 10.10.37.92 >>>>> node mds91 mds92 >>>>> respawn root /usr/lib64/heartbeat/pingd -m 100 -d 1s -a pingd >>>>> ping 10.10.37.100 >>>>> >>>>> The content of */var/lib/heartbeat/crm/cib.xml*: >>>>> Please see the attachment. >>>>> >>>>> 2 nodes in active passive mode. >>>>> 'restart network service ' on active node causes brainsplit. Please >> see >>>> the >>>>> ha-log on the attachment. >>>>> >>>>> Thanks. >>>> >>>> _______________________________________________ >>>> 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 >> > _______________________________________________ > 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
