[Samba] runaway samba, non-responsive daemons

[Gary Browning]

Hello,

I am wondering if someone may have experienced this or might know what is 
happening. First, let me lay out the ground work so you understand our 
environment.  We have several large servers (dual processors, gig of ram, 
300+gig drive space) that are running redhat versions 7.0-7.3.  Some of the 
severs are running samba 2.2.5, 2.2.5 (with spoolss patches), 2.2.4.  We 
have a samba PDC and BDC.

We have used samba for a number of years without problems.  We have several 
hundred machines that used to be able to get their personal drive letter 
mapped from a single samba server.  Granted, everything was flawless under 
samba 2.0.x and actually, everything worked under samba 2.2.x until we 
introduced Windows XP into the blend.. or at least that's what I believe 
where things started to go wrong.

Under windows XP, authentication works flawlessly; however, at times the 
mapping of drive letters hangs.  On the server end we see several smbd 
processes that are owned by root.  They eventually accumulate to over 1000 
causing the number of open files on the server to climb.  Since the samba 
daemons are not responsive we try to shutdown samba using the init.d 
script, which fails.  A kill -HUP fails and it requires a hard kill to stop 
the smbd daemons so that we can refire it up.  Once samba is running again, 
all it well until another high traffic time.  We have tried leaving samba 
alone; however, it does not recover which is why is we to restart 
it.  Also, our windows xp clients hang for minute waiting for the mappings 
to time out.

As I said before, the PDC and BDC work flawlessly and they are not the ones 
that are dying.  Any of the servers doing file services or print services 
have issues.  We used to be able to keep 250+ user connections at a single 
time, but not anymore.

Here is some additional information.  We were using roaming profiles under 
XP; however, as a debugging method we turned those off.  The version of 
redhat (7.0-7.3) has no bearing on the samba processes as we have had all 
file serving samba servers have issues.

Any ideas, thoughts, commiserations would be appreciated.

Thanks,
Gary

(a copy of one of the member file server's smb.conf is listed below)

[global]
    workgroup = BURNING
    netbios name = EMBERS
    server string = Data Server
    hosts allow = (trimmed .. in fashion of x.x.0.0/255.255.0.0)
;   utmp = yes
    deadtime = 15
    printcap name = /etc/printcap
    load printers = no
    printing = bsd
    log file = /var/log/samba/log.%m
    max log size = 50000
    log level = 1
    security = domain
    password server = *
    encrypt passwords = yes
    socket options = TCP_NODELAY SO_RCVBUF=8192 SO_SNDBUF=8192
    interfaces = eth0
    local master = no
    os level = 33
    domain master = no
    preferred master = no
    domain logons = no
    name resolve order = wins hosts bcast
    wins support = no
    wins server = (trimmed .. ip address of our wins server was here)
    wins proxy = yes
    dns proxy = no
    default case = lower
    max open files = 30000
    time server = Yes

#============================ Share Definitions ==============================
[homes]
    comment = Home Directories
    browseable = no
    writable = yes
    hide files = /desktop.ini/