[Veritas-ha] Is there a resource-type for nfs mounts?

[Dylan Northrup]

On Mon, 29 Oct 2001, Barry Ringuet wrote:

:=Pretty sure there isn't but thought I'd ask, Is there a resouce-type for
:=nfs mounts?
:=Obviously I see the 'mount' type, but it requires a block device.
:=Am trying to use an nfs share, from a non-clustered machine, as an
:=apache docroot.

This is a tricky thing to do.  Starting and stopping the resource are
trivial to do.  But monitoring becomes tricky.  You need to do some I/O to
the mount (preferrably, read information from a file, write information to a
file, then read information again to make sure it changed or some such).
The problem is when I/O to a mount point is horked (i.e. the device sharing
the resource is down) your test will block on I/O indefinitely.  You'll have
to write something that will timeout the I/O access (which I'm not sure is
possible, but am glad to be proven wrong).

Considering you're using a single, non-clustered machine as your NFS
machine, you could just skip the test and just write a start and stop
script.  If you failover from one machine to another, it will still fail
when it tries to mount the unreachable NFS machine so there's no use in
failing over in the first place because of a failure of the NFS mount.

When we were doing this for a major storage appliance manufacturer we were
told to assume the NFS mounts wouldn't go away (and the storage appliances
would be clustered to take care of any failure).  So, we just worried about
how to properly mount and unmount the file systems and did rudimentary
testing with the understanding that the I/O accesses would block at most for
a minute or so while the appliances failed from one to the other.

Hope this helps.

--
Dylan Northrup <*> northrup at loudcloud.com <*> Evil Dylan, the Unix Dylan
"Harder to work, harder to strive, hard to be glad to be alive, but it's
 really worth it if you give it a try." - Cowboy Mouth, 'Easy'