On Fri, 2010-01-08 at 16:54 -0700, hj lee wrote: > Hi, > > My simple two-node cluster is configured with two ring interfaces > (ring 0 and ring 1). If I disconnect one of cable at one of nodes, > then openais enters to GATHER mode and returns to OPERATIONAL mode. > This takes 200 - 300 msec. In GATHER mode, the openais does not seem > to send and receive any message from Pacemaker like messages for > updating cib files. > > What is the reason behind of this behavior while the openais is still > connected with other ring interface? Is it possible to disable this > behavior? > > In my configuration, the eth0 is used both for openais and pingd > monitoring. So when eth0 is disconnected, pingd fails and also openais > enters GATHER mode at the same time. The pingd failure trigger the > resource migration, but this migration is delayed by 200 - 300 msec. I > want to get rid of this delay when pingd fails. >
This seems quite odd. Gather should not be entered unless the network interface is actually changed. It may be that you are running a tool such as network manager which is downing the interface. Running network manager with corosync is likely to provide bad results because network manager destroys interfaces within the kernel. Please report back if this is your issue. Regards -steve > Thanks > hj > _______________________________________________ > Openais mailing list > [email protected] > https://lists.linux-foundation.org/mailman/listinfo/openais _______________________________________________ Openais mailing list [email protected] https://lists.linux-foundation.org/mailman/listinfo/openais
