|
We are using Samba 2.2.4
(compiled from the web download) running on a Compaq ES-40 cluster
(2-nodes) under tru64 v5.1.
Our problem is two-fold and
any pointers on solving either of these would be greatly
appreciated.
Shares disappear
for no reason and no traces left in the logs. This can happen
once or twice a day. Simply restarting Samba resolves the problem, but is
embarrassing as the users often find it before the IT staff. I am
considering using modifying one of the "torture programs" to act as a "keep
alive". Any thoughts??
NMBD hangs in
"Uninterruptible" state. Originally when this was happening, we
noticed that NMBD was also listening for UDP packets on the cluster "memory
channel" interface. I have since modified my smb.conf to look only at the
appropriate tu0 interface of the system, but we still have NMBD processes that
hang in the "U" state. There is then a one to one correlation between
these hung processes (they cannot be killed by anything other than re-booting
the system) and process that appear to be waiting for UDP packets on ports 137
and 138. Have we possibly missed something in setting up NMBD for the
cluster environment? Any suggestions on patches that could be made so that
we could at least kill these processes with and unusual signal (perhaps a
SIGUSR1 or SIGUSR2). Again any help in resolving this problem would be
greatly appreciated.
|
[Samba] Samba 2.2.4 running in Tru64/v5.1
William Wachob [EDCC / BannerIT] Tue, 18 Jun 2002 10:30:44 -0700
