Problems with Samba and Solaris 2.6 (and STATUS..LCK)

[Tom Lieuallen]

We are also running 1.9.17p4 on an Ultra 450, 2 procs, 512 MB, Solaris
2.6.  We have had several problems that sound like the same as reported
below.  We usually have around 100 smbd processes running.
Occasionally, the machine just locks up (or seems to).  We sent the
system crash files to sun -- and they are initially saying there were
316 samba processes running at that time.  It looks like it ran away
and started paging its life away.

I'm planning to cut over to 1.9.18p3 in the next day or two, but
I have reservations.

We have another machine (ultra 140, solaris 2.6) that was running
1.9.17p2.  When I upgraded it to 1.9.18p3, it just didn't work
for a number of hours.  We are doing 'security=server' and passing
the authentication on to our NT PDC.  The error messages sounded
like the NT server was only giving 'guest' authentication, which I
guess means that normal encrypted authentication had failed.

When I was using smbstatus, It kept complaining about there being
no STATUS..LCK.  I tried killing the samba processes, etc. etc.
I never touched the config file.  After a while (quite a while),
it just started working.

Could there be some NT issue at work here?  Maybe our NT PDC thinks
that we've changed hardware -- and it needs to timeout?

I would really like to know before I upgrade our main server.  I
can't afford to let that one sit out in the cold for an unspecified
amount of time.  I had a very similar experience when I ugpraded the
first machine from 1.9.17 to 1.9.17p2 (6+ months ago).

Here's the global portion of our smb.conf, if it is any help

[global]
   wins support = no
   wins server = foo.engr.orst.edu
   debug level = 1
   security = server
   password server = ntpdc
   preserve case = yes
   short preserve case = yes
   case sensitive = no
   log file = /private/samba/var/log.smb
   locking = yes
   lock directory = /private/samba/var/locks
   share modes = yes
   guest account = samba
   workgroup = ENGINEERING
   server string = Engineering Unix gateway - %h
   domain master = no
   local master = no
   preferred master = no
   os level = 0
   fake oplocks = yes


Thanks for any help

Tom Lieuallen
Oregon State University

> Hi all,
> 
> We are running samba 1.9.17p4 on two Sun Ultra 450, Solaris 2.6
> with all recommended patches applied.
> Both macines have two CPUs and are working just fine - no problems
> with going from 2.5.1 to 2.6.
> 
> Samba binaries were compiled under solaris 2.6 using gcc 2.7.2.3
> I got the compiler as a 2.6 package from
> http://www.mbp.duke.edu:80/christensen/sitesNG.html
> 
> Olga
> 
> You wrote
> | We're having problems running Samba on a dual-processor Sparcserver 10
> | running Solaris 2.6. The problem is that samba seems to spawn multiple
> | processes for each connection and generally consume more and more
> | resources until the system runs out of memory and swap, or just 
> | generally becomes completely unresponsive.
> |
> | We have seen load averages of 500+ on this machine as a result.