[Samba] 3.x & Solaris 8 lockups

[Tom Schaefer]

Is anybody experiencing what I have twice now -- Samba 3.x running fine on
Sparc/Solaris 8 for a week or more then one day out of the blue it'll just
go all haywire?  So haywire that in both my instances of this I had to
change Samba versions to get the servers back to life.

Long boring story short on any useful detail below...

Recently I had to go from 2.x to 3.x series since there aren't going to be
any more patches released for 2.x.

I started in July with 3.0.2a obtained as a package from sunfreeware.com
and it worked flawlessly for about a week.  Then one day the smbd
processes kept freezing and new ones would get spawned for the same
already connected user.  You had to use a -9 to get them all shut down.  I
needed to get it back in action quick since it is a production server so I
couldn't really spend time troubleshooting.  I had recently compiled 3.0.5
myself so I swapped that in, using the exact same smb.conf and what not. 
Blamed it on never knowing exactly what you're getting from sunfreeware
and/or older version of samba.  Never had a problem with that server
since.

A little over a week ago I migrated that server from 3.0.5 and 5 other
servers from 2.x to version 3.0.7 which I compiled myself.  I installed
that same compile on all 6 servers.  Everythings been great for the past
10 days or so since the upgrade then today all of a sudden one of the
servers is acting like the one other one did back on me in July.  Horrible
performance from end users perspective, smbds not responding so new ones
being launched until the server ran out of swap space, even after
rebooting that server same problem today, stopped and started samba a few
times, have to use pkill -9 smbd to get rid of them all, that server just
refuses to run 3.0.7 today so I had to revert back to the 2.2.8a that I'd
upgraded from.  I really can't provide any detail since in both cases it
was extremely urgent that I just get them working again ASAP and didn't
have time to experiment and turn up the log level and what not.

Outside of these two instances though its been great on all 6 servers and
a couple test ones.

Tom Schaefer