Problems with AIX

Michael E Osborne mosborne at
Wed Oct 13 01:24:32 GMT 1999

We've been running Samba under AIX since 1.9.x and with the exception of a
few glitches, it has been rock solid. Unfortunately, since we have moved to
AIX 4.3.2 and Samba 2.04+, (currently 2.0.5a) stability has dropped
dramatically. I have been unable to figure out what is going on. We have
normally from 75-200 users logged on at any given moment and the
connections are fairly active.

Samba will run anywhere from a few hours to several days until it seems to
go berzerk. New samba processes run for a bit and then die, to be replaced
by another, which dies, and so on. When this happens running smbstatus will
result usually result in a segmentation fault when it tries to list the
locks. Sometimes it gets into a loop and simply lists processes over and
over again. Killing all Samba processes and restarting will let it run for
another stretch. When the Smaba processes die I sometimes get a Signal 11
received recorded in the logs, but not always.

We also get the following error in our logs:

read_socket_data: recv failure for 4. Error = Connection reset by peer.

Looking through Dejanews and these lists I have seen this error reported on
multiple platforms, usually associated with 2.0.5a.

I have tried running Samba from inetd and as -D. Both options act the same.
I have recompiled with different options to no effect. The AIX 4.3.2
binaries disributed by Bull also have the same behavior.

We have a mix of WFW 3.11, NT 4.0 (SP4, SP5), and OS/2 3.0. We usually
      notice the problem first under NT.

Any ideas what might be going on here?

One strangeness that we noticed when we first moved from 4.[12].x to 4.3.2
      was unless the "max open files" parameter is set in smb.conf, the
      smbd processes would grow to over 12MB in size.

Michael Osborne

Raytheon Demilitarization Co

Johnston Atoll, Central Pacific Ocean :-)

Disclaimer: All opinions are mine!

More information about the samba-technical mailing list