NFS / automount hangs on Solaris 10 SunOS oats 5.10 Generic_125100-07 sun4u

Question

We are running Solaris 10 125100-07 on a SunFire V440.
We host computer services for another department on this host.
The server acts as both an NFS server and an NFS client. We do a lot of cross-mounting in our environment due to the lack of large centralized storage.
This particular machine automounts shares from our Solaris 8, 9 and 10 servers and possibly a few Suse servers, and a NetApp owned by another department.
Every day and a half or so, something happens to either NFS or the automounter, and the machine 'hangs' in that we have NFS requests that never get filled.
Restarting either autofs or the nfs client, or both does not seem to help. The only fix we have found so far is a reboot.
We have throttled NFS_CLIENT_VERSMAX to 3 and set AUTOMOUNT_VERBOSE and AUTOMOUNTD_VERBOSE both to TRUE.
The general failures we see look like this:
May 17 04:44:15 oats nfs: [ID 333984 kern.notice] NFS server grieg not respondin
g still trying
May 17 04:44:15 oats last message repeated 1 time
May 17 04:46:15 oats automountd[392]: [ID 196269 daemon.error] dupreq_nonidemp:
duplicate request in progress
May 17 04:49:57 oats last message repeated 4 times
May 17 04:52:15 oats automountd[392]: [ID 196269 daemon.error] dupreq_nonidemp:
duplicate request in progress
The dupreq_nonidemp messages repeat until we reboot.
The nfs server that does not respond varies from day to day. No other computer on our network complains about it.
This started after we installed 125100-05, and has continued through 125100-07.
As far as I can tell we are up to date on patches:
root@oats # smpatch analyze
No patches required.
Any thoughts?
I was thinking I might back out 125100-XX.
dal
e

Answer 1

Interesting,
we have the same problem on SunOS Generic_118833-36 on V440, V445, V240, V245, but not on an old E420, and not on a T2000.
We upgraded them with patch cluster to Generic_138888-07, but it has not changed anything:
E420 and T2000 are stable, the Fire-Vxxx have problem, each has its own "problem interval": one system 1.5 days, two systems 10 days, one every 3 months.
The problem is: NFS/TCP is suddenly dead. Then all NFS mounts, e.g. a manual mount, is hung with "NFS server not responding". A reboot helps.
We have added proto=udp to the mount options. No more problems since.
Permanent mounts, without Automounter, would certainly work, too.
But IMHO there is nothing wrong with Automounter itself, but the hanging occurs e.g. after 4000 NFS mounts, and Automounter is the enabler for that.