Hello All

We're using a couple of NFS-mounts to share software-repositories (patches, servicepacks etc.) across two datacenters but it seems that this is not the most fault tolerant solution in the world.

If the network between the two locations goes down the server acting as NFS-client starts building up load to a point where it becomes unusable. The CPU is more or less idle but the average load goes up from the second the link between the servers are broken. Restoring the connection may or may not solve the problem and reduce the load again.

It's usually impossible to unmount the NFS mount on the client side once the connection to the server is gone - sometimes a restart of netfs may help but most of the times it just hangs while trying to restart.

The real killer is the fact that we cannot even reboot the server remotely - it'll hang, trying to unmount the stale NFS mount and must be reset manually. That is just SO wonderful in the middle of the night....NOT!

Does anyone know of a way to make this crap a bit more foregiving or is NFS just not suited for sharing mounts via networks....? ;-)

There are two types of NFS mounts, hard and soft.

Hard says: "I don't care fio teh server is down I will keep trying until it comes back!"

This is the scenario you're experiencing. There is an interrupt option but you have to catch the "stale" mount as it happens.

Soft says: "You down? That's OK, I'll leave too."

This requires some manual intervention once the server is accessible again, but the mounting server won't be "hosed".


The best of both worlds is automounter. People of had problems with it, but I haven't and use it in a couple of websphere environments.

Ah - a brand new world is opening up! :)

Just tried with "soft,intr" in fstab and liked the result very much indeed. Thanks a lot.