2009/4/22 Ralph Zukeb <ralphzukeb@xxxxxxxxx>: > 2009/4/22 Mark Watts <m.watts@xxxxxxxxxxxxxxxx>: >> >> On Wednesday 22 April 2009 11:14:49 Ralph Zukeb wrote: >>> 2009/4/22 Mrugesh Karnik <mrugeshkarnik@xxxxxxxxx>: >>> > On Wednesday 22 Apr 2009 15:25:32 Ralph Zukeb wrote: >>> >> Hello again! Thanks for the tips so far. >>> >> >>> >> How can I ensure that the node holding the services and resources in a >>> >> two node cluster keeps holding them if the two nodes cannot see each >>> >> other? >>> >> >>> >> I am not using a quorum disk. >>> > >>> > <cman two_nodes="1" expected_votes="1"/> >>> > >>> > Make sure you configure fencing properly though. Also, getting rid of the >>> > two_node attribute requires a full cluster restart. >>> > >>> > Mrugesh >>> >>> I tried this, but when I killed the switch for the cluster traffic, >>> BOTH nodes got fenced! Can I avoid this? >> >> Don't kill the switch like that - you're causing a split brain. >> Use two switches and multiple interfaces (bonding) to provide network >> redundancy. >> >> Mark. > > I have bonding, but both network paths go over the same physical > cable. I want to simulate a cable cut or a switch dying along this > stretch of cable. > Maybe I can specify a different fencing delay per host? -- Linux-cluster mailing list Linux-cluster@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-cluster
