Problems with NetApp filer connected to Sun V440 via NFS when using Oracle

  • From: stephen booth <stephenbooth.uk@xxxxxxxxx>
  • To: "oracle-l-freelists.org" <oracle-l@xxxxxxxxxxxxx>
  • Date: Tue, 4 Jan 2005 18:10:43 +0000

We have a bit of a wierd problem using a NetApp filer  to hold the
datafiles for an Oracle database (9.2.0.4) for a Sunfire V440 (Solaris
9 patched to the most recent patchset for Dec 04).  I wondered if
anyone else had come accross anything similar, given that (according
to NetApp) their filers are used so commonly with Oracle.

The filesystem on the filer is exported and mounted by the V440 via
NFS (as per reccomendation by NetApp).  With Oracle down the mount
stays mounted indefinately (we've tried it for 3 weeks, about half
with no traffic accross the link and the remainder with a file being
repeatedly copied accross the link from the filer to the server and
back again).

With Oracle up the NFS connection seems to drop out non-fatally for a
few minutes.  Oracle just freezes, no messages in the alert log but we
do get a few messages in the messages file (OS log) such as:

Dec 22 14:18:22 XXXXXXXXX nfs: [ID 333984 kern.notice] NFS server
filer not responding still trying
Dec 22 14:21:53 XXXXXXXXX last message repeated 2 times
Dec 22 14:29:17 XXXXXXXXX nfs: [ID 563706 kern.notice] NFS server filer ok
Dec 22 14:29:17 XXXXXXXXX last message repeated 1 time

After a  while (could a few minutes or quarter of an hour, it varies)
the connection comes back and oracle continues on as if nothing had
happened.

I've tried looking in v$waitstat to see if it is anything particular
but nothing jumps out as having massively higher wait times than
everything else.  Although given that the connection drops out for
several minutes it's likely that everything will get hit badly.

Has anyone run accross this or something similar?  If so, how did you solve it?

Is there anything Oracle might be doing to cause this?  The only thing
that comes to mind is that Oracle is doing random access where as the
copying the file back and forth is doing sequential access.

Thanks

Stephen
-- 
It's better to ask a silly question than to make a silly assumption.
--
//www.freelists.org/webpage/oracle-l

Other related posts: