Mailing List Archive

It sounds like the NFS locks aren't correctly being recognized by
the server that fails over. To test it, on the primary create a file in
/var/lib/nfs. ie something like

$ touch /var/lib/nfs/failover.test.file

Then fail over and see if the failover.test.file exists in the
/var/lib/nfs directory of the secondary machine.

If that file isn't there, something went wrong when you created
all the symlinks and that's why you've got stale file handles upon
failover.

Another approach I've heard off to make failover NFS easier is by
using stateless NFS. I've never set this up but from my understanding
stateless nfs makes it so the NFS servers does not use lock files. It
sacrifices some speed, but makes failing over with NFS as easy as a web
server since you don't have to worry about NFS lock files getting
replicated.

On Thu, 22 Nov 2001, Espen Myrland wrote:

>
>
>
>
> Hello.
>
> I have followed the example of how to make a failover nfs-
> server. My problem is that when node A fails, node B is coming up,
> but the nfs-clients nfs-handles become staled,
> and I have to umount, and mount clients again.
>
> I'm running depian potato, with perc-raid on both
> nodes, which is mirrored with drbd.
>
>
>
> Anyone having the same problem, or any tips?
>
> Maybe I should post this somwhere else?
>
>
> regards,
>
> --
> espen
>
>
>
> _______________________________________________
> DRBD-devel mailing list
> DRBD-devel@example.com
> https://lists.sourceforge.net/lists/listinfo/drbd-devel
>

Espen Myrland <espenmyr@example.com> wrote:
>
> Hello.
>
> I have followed the example of how to make a failover nfs-
> server. My problem is that when node A fails, node B is coming up,
> but the nfs-clients nfs-handles become staled,
> and I have to umount, and mount clients again.
>
> I'm running depian potato, with perc-raid on both
> nodes, which is mirrored with drbd.
>
> Anyone having the same problem, or any tips?
>
> Maybe I should post this somwhere else?

You have to put /var/lib/nfs also on a shared partition/DRBD-partition...

Juri

--
Juri Haberland <juri@example.com>