Hugo Lafargue wrote:
> Hi,
>
> I just installed a cluster of 2 nodes to provide High Availability NFS/Samba
> service with Hearbeat (0.4.9.1).
> The exported filesystem is XFS.
> The kernel is 2.4.18 patched with xfs-1.1
>
> When I stop Heartbeat daemon on the nodeA, the shared ip swithes to the
> nodeB, and samba clients still access the files.
>
> But, I'm now observing that after failover, nfs clients are unable to use
> the filesystem exported by the backup server until I unmount it and mount
> it again. Otherwise, I get "stale NFS file handle" messages.
>
> Is this a known issue? Is there a way to fix it ?
>
> I find on google that there was a known issue with exporting ReiserFS over
> NFS, but it seems to be
There are some things that you need to care about to make this work:
The exported file systems must have the same major/minor device number on
both nodes. Also /var/lib/nfs should be on a shared file system used by
both nodes. And you might need to run the latest nfs-utils (or maybe even
the cvs version) to be able to tell the nfsd that it should use the
cluster name (the DNS name corresponding to the shared IP) instead of the
node name.
Search the archives of the linux-ha and linux-ha-dev lists at
http://marc.theaimsgroup.com/?l=linux-ha and
http://marc.theaimsgroup.com/?l=linux-ha-dev for more details.
[any comments, Ragnar?]
Cheers,
Juri
|