Hi!
I just searched the archive for a solution to our problem:
We've got two machines clustered with a shared storage so there's
always one clusternode active and the other waiting for the active to
fail (hot stand-by configuration). Both clusternodes have the same
configuration: We're running Samba 2.2.1a on Red Hat Linux 7.1, / (the
node's local disk) and /shares (on the shared storage) as ReiserFS,
and Kernel 2.4.9, since we ran occasionally into a kernel bug in
memory management with Red Hat's 2.4.2-2 factory kernel.
[root@stan logrotate.d]# uname -a
Linux stan 2.4.9 #1 SMP Thu Sep 27 15:59:08 CEST 2001 i686 unknown
We have about 500 WinNT 4.0 SP6a clients connection to the cluster.
Samba 2.2.1a and 2.2.0a have been running finde for a couple of weeks
but since a few days, clients can't connect anymore. This happens
regardless which of our clusternodes is active so I dare to say that
we can assume that it's *not* faulty hardware.
We found the below message in the client logfiles (smb.conf: logfile
/shares/sambalog/log.%m). I just took a look at the file
/var/lock/samba/sessionid.tdb and it looks strange to me. Excerpt from
our /var/lock/samba/sessionid.tdb:
[2001/10/04 07:30:37, 0] tdb/tdbutil.c:tdb_log(342)
tdb(/var/lock/samba/sessionid.tdb): tdb_oob len 875767935 beyond eof at 797816
[2001/10/04 07:30:37, 0] tdb/tdbutil.c:tdb_log(342)
tdb(/var/lock/samba/sessionid.tdb): tdb_oob len 808465011 beyond eof at 797816
[2001/10/04 07:30:37, 0] tdb/tdbutil.c:tdb_log(342)
tdb(/var/lock/samba/sessionid.tdb): tdb_oob len 909195343 beyond eof at 797816
[2001/10/04 07:30:37, 0] tdb/tdbutil.c:tdb_log(342)
tdb(/var/lock/samba/sessionid.tdb): tdb_oob len 808465011 beyond eof at 797816
I'd say this looks like a logfile, not like a database. Am I right?
Stopping Samba and deleting the contents of the /var/lock/samba-dir
helps but doesn't cure the problem. Sooner or later we have the same
situation.
As I said, I searched the archive. I've found similar situations with
different setups (Samba on HP-UX, WIN2K clients, etc.) Our Samba runs
on Intel Hardware and our clients are WinNT. Maybe there's some Win9x
between, but there's no known Win2K which I will check again to make
it sure.
Samba has been compiled with the Red Hat gcc:
[root@stan logrotate.d]# gcc -v
Reading specs from /usr/lib/gcc-lib/i386-redhat-linux/2.96/specs
gcc version 2.96 20000731 (Red Hat Linux 7.1 2.96-81)
Here's my configure options:
./configure \
--prefix=/usr/local/samba \
--exec-prefix=/usr/local/samba \
--bindir=/usr/local/bin \
--sbindir=/usr/local/sbin \
--libexecdir=/usr/local/libexec \
--datadir=/usr/local/samba/share \
--sysconfdir=/shares/etcsmb \
--sharedstatedir=/usr/local/samba \
--localstatedir=/shares/sambalog \
--libdir=/usr/local/lib \
--includedir=/usr/include \
--infodir=/usr/local/samba/info \
--mandir=/usr/local/man \
--with-pam \
--with-privatedir=/shares/etcsmb \
--with-lockdir=/var/lock/samba \
--with-swatdir=/usr/local/samba/swat \
--without-sambabook
This is smb.conf:
[global]
workgroup = SR
netbios name = USERHOMES
server string = Samba-Cluster %v
load printers = no
log file = /shares/sambalog/log.%m
max log size = 0
security = domain
encrypt passwords = yes
name resolve order = wins host lmhosts bcast
password server = 192.168.1.3
# password server = 192.168.1.2
# socket options = TCP_NODELAY
interfaces = 192.168.0.69/255.255.255.0
bind interfaces only = yes
local master = no
os level = 33
domain master = no
preferred master = no
wins server = 192.168.1.2
dns proxy = no
create mask = 0777
directory mask = 0777
add user script = /shares/etcsmb/smb_useradd.pl %u
kernel oplocks = no
oplocks = no
#============================ Share Definitions =============================
[homes]
comment = Home-Verzeichnis %u
browseable = no
writable = yes
guest ok = no
valid users = %S administrator
I've got absolutely no idea what's the cause of this error.
As this is an server already in production, any hints (use a different
gcc), workarounds (some configure options) and other solutions
(downgrade do Samba 2.2.0a?) will be gratefully accepted.
Best regards,
Christian Seip
----------------------------------------------------------------------
*** Achtung: Neue Adresse! *** Achtung: Neue Adresse! *** Achtung: Neu
Christian Seip Tel (06386) 404 304
Diplom-Informatiker (FH) Fax (06386) 404 305
Petschwiesen 19 Mobil (0171) 61 70 886
66904 Br?cken/Pfalz E-Mail cseip@cse-consult.de
WWW http://www.cse-consult.de/