On Thu, Dec 18, 2008 at 19:26, Peter Niessen <[email protected]> wrote: > Hello, > > first, thanks to the heartbeat girls and guys for providing a piece of > easy administrable software! > > I'm working on an active/active NFS server, in which two servers have > access to two common storage arrays. I.e. we have servers nfs1 and > nfs2 which both can see /dev/sdb1 and /dev/sdc1. In normal operation, > nfs1 should serve /dev/sdb1 as /scratch1 and nfs2 should /dev/sdc1 as > /scratch2. I'm using heartbeat 2.1.3. > > With the help of > > http://web.archive.org/web/20070630065713/http://chilli.linuxmds.com/~mschilli/NFS/active-active-nfs.html > > I created identical /etc/fstab files on nfs1 and nfs2 where /scratch1 > and /scratch2 are listed with the "noauto" option: > > /dev/system/scratch1 /scratch1 xfs rw,suid,dev,exec,noauto,nouser,async 1 2 > /dev/system/scratch2 /scratch2 xfs rw,suid,dev,exec,noauto,nouser,async 1 2 > > > The /etc/exports on each machine list /scratch1, /scratch2 as > exportable. > > In a first approach, I put together two groups (ordered, colocated) which > provide an ip-address (ocf IPAddr2) (10.0.0.1, 10.0.0.2) and the > mountpoint (ocf Filesystem), one group for each server. > I also created two locations so that the servers run on their defaults. > > The nfsserver is created as a cloned resource, one instance running on > each file server. However, when failover occurs, the surviving clone > of the nfsserver isn't restarted,
Right. This is by design. What the other instance will do, if you configure it to, is receive a notification that the other side stopped. So what you need to do is configure notify=true for the nfsserver resource and modify the script to restart if it is told that a peer went away. > which results in "stale nfs handle". > > How can I tell the nfsserver to restart on failover? > > Cheers & thanks, Peter. > > PS: Why did I not put the nfs servers into the group resources? > > When the nfsserver is part of the group, the fail-over works great: > The ip-adress is transfered, the partitions are mounted and the nfsserver > is restarted, so the nfs clients keep seeing the disks. > > For example, if nfs2 fails, 10.0.0.2 moves over to nfs1, and /scratch2 > is mounted on nfs1. > > Now, fail-back has a problem: When the relocated source is migrated > back to the re-started node (nfs2), it stops the nfsserver on the failover > node (nfs1) and the clients get in trouble because the /scratch1 > filesystem disappears. > > -- > Peter Niessen > FZJ > Tel.: (+49)2461/61-1753 > > > > ------------------------------------------------------------------- > ------------------------------------------------------------------- > Forschungszentrum Juelich GmbH > 52425 Juelich > > Sitz der Gesellschaft: Juelich > Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498 > Vorsitzende des Aufsichtsrats: MinDir'in Baerbel Brumme-Bothe > Geschaeftsfuehrung: Prof. Dr. Achim Bachem (Vorsitzender), > Dr. Ulrich Krafft (stellv. Vorsitzender), Prof. Dr. Harald Bolt, > Dr. Sebastian M. Schmidt > ------------------------------------------------------------------- > ------------------------------------------------------------------- > _______________________________________________ > 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
