Original question is found at
http://www.uio.no/~kjell/du/malloc.txt
I forgot to tell you our site has a lot of machines running DU v4.0D and
no other machine have reported the same problem.
I think this may be because this servers is the one with most users and
most NFS-exported disks.
Thanks to all these people for their immidiate answers:
Sean O'Connell <sean_at_stat.Duke.EDU>:
-----------------------------------------------------------------
I believe that there is a fairly nasty bug in the nfs server
code under 4.0D upto and including patch 2.
C.Ruhnke <i769646_at_smrs013a.mdc.com>
-----------------------------------------------------------------
I had a similar problem on my dual-system AS 8200 TruCluster DU V4.0D.
First, have you installed the mondo-patch Kit #2?
If so, check with your software support contacts. There is an unpublished
patch that has modifications to nfs services and file services to better
utilize memory buffers. The latest version I have received is identified
as v40d_bl10_moiz_kohari_patches and was dated about Nov. 7th.
Mitch Bertone <mbertone_at_gtech.com>
-----------------------------------------------------------------
Try reading this summary, seems to have some useful information about
same type of problem. I hope it helps,
http://www.ornl.gov/its/archives/mailing-lists/alpha-osf-managers/1998/08/msg00210.html
Erik Piip <piip_at_merl.com>
-----------------------------------------------------------------
There are a BUNCH of vm and nfs fixes for 4.0D. I saw this behavure, where
the system would stop.
Start with the latest patch kit, and then ask services to get you the most
recent patches....
Jason Orendorf <orendorf_at_alf.dec.com>
-----------------------------------------------------------------
Many similar problems have been reported and a patch has been developed by
Digital UNIX Support Engineering. The fix should be in the
soon-to-be-released patch kit #3 for 4.0D. If the problem needs to be
fixed at your site before kit #3 arrives, contact your local CSC to obtain
a custom Engineering patch to prevent the problem.
Mike D Cross <crossmd_at_mh.uk.sbphrd.com>:
-----------------------------------------------------------------
There is a bug in the nfs drivers which eats memory, the more users
use the file server the quicker you run out.
There is a patch set available on request for most recent DU releases, you
will probably have to log a call with dec to get the patch set you need.
Kjell Andresen Systems administrator, University of Oslo, Norway
Center for Information Technology Services and
Department of Geophysics
Received on Fri Nov 20 1998 - 08:40:52 NZDT