[Samba] Linux+samba problems

[Voetelink D.]

Hi,

A few weeks ago we replaced our Alpha with Tru64 UNIX and samba with a Compaq Proliant
ML370, running RedHat 7.3 and Samba 2.2.7.

We have about 1000 clients connecting to the samba shares, with max. about 500 at any time.
The clients are running windows 95, 98, NT4 and 2000.

Now the problem: from time to time the server 'hangs'. It's still pingable. If I use telnet
I do get a connection, but it doesn't continue to log me in. Even when connecting to port 22
(ssh) I get a connection, but the ssh server doesn't announce its version to me (as it does otherwise).
Also at the console I cannot so anything with the server, apart from powercycling it.

This happens about twice a day now, except for the weekends.

We analyzed the output from sar. All we could find was that 'totsck' (total sockets) is rising, and
reaching 900/1000 before the hang. ( approaching 1024 ??? ).

We tried running samba 2.2.7a, but this didn't help. Also, the system has all updates provided
by Red Hat installed.

Has anyone seen this behaviour or know a solution?

I'm not really sure samba is the cause, but it is one of the major tasks of the system.


Please help!

Dennis