Brad Filipek escribió: > Should I manually mount the GFS file system individually on both nodes > by running “mount –t gfs /dev/sdb2 /mountpoint”? Or can I add the GFS Hello, Brad: I'm currently running a two-node mail cluster using GFS, and while I'm able to do mount -t, I keep the filesystem entries in /etc/fstab so I keep standard options such as noatime,nodev,nosuid,noexec as well as the mount on boot option. If you add _netdev to your fstab entries, you'll make sure that your partitions are mounted on boot, which is generally a Good Thing (TM). Make sure to configure your cluster in the two-node mode, which is done by adding the twonode="1" and expectedvotes="1" arguments to your cman declaration in your cluster.conf. This will make the cluster quorate even with one node up. If I understand correctly, when one of your nodes reboot, if you're using a sane initscripts configuration, your partitions should shutdown as well as other daemons such as fenced, clvmd, cmand and ccsd. This will tell your other node that its partner is going down and it will keep working just as usual. I haven't tried this in my actual configuration, but I can start and stop cman and the other node keeps working OK. However, if one of the nodes goes down suddenly, the other node will stall its writing to your clustered filesystem until you acknowledge the fencing (that is, assuring that the shutdown node won't write to the shared storage) either by using the manual method (fence_ack_manual) or by interacting with hardware such as a UPS. The cluster documentation should provide enough information on available methods and configuration. HTH, Jose -- Linux-cluster mailing list Linux-cluster@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-cluster
