Jeremy Allison wrote:
On Sat, Nov 08, 2008 at 06:28:42PM +0100, Marco van Putten wrote:
Hi,
We're running Samba (3.0.28-1.el5_2.1) on a Redhat (5.2) server using
winbind to authenticate against a Windows AD.
This all worked fine for a while. But now we have the problem that
(some) users coming from Linux clients cause an excessive amount of smbd
processes. We had 1 user who had more than 1700 smbd processes all
coming from the same client. While on the client side there where only a
few cifs processes running.
All these processes are not responding to a normal kill. They have no
more connection to the client what so ever and have to be killed with a
kill -9.
Attach to these processes using strace -p <pid> and find out what
they're up to. Also try attaching with gdb <attach> command and
get a back trace. Don't forget to load the symbol tables also.
This should give some clue as to what is going on.
Jeremy.
Thanks for your reply Jeremy.
These are the outputs I got from gdb and strace. It looks like there is
not very much going on... Hopefully it makes some sense to you.
# gdb smbd 32717
(gdb) bt
#0 0x00002b55a75a9ddf in fcntl () from /lib64/libpthread.so.0
#1 0x00002b55a5e60256 in tdb_brlock () from /usr/sbin/smbd
#2 0x00002b55a5e60065 in tdb_lock () from /usr/sbin/smbd
#3 0x00002b55a5e601c0 in tdb_chainlock () from /usr/sbin/smbd
#4 0x00002b55a5dfa380 in is_locked () from /usr/sbin/smbd
#5 0x00002b55a5df9e38 in is_locked () from /usr/sbin/smbd
#6 0x00002b55a5cc5c11 in reply_write_and_X () from /usr/sbin/smbd
#7 0x00002b55a5cfda14 in schedule_deferred_open_smb_message () from
/usr/sbin/smbd
#8 0x00002b55a5cfe9c1 in smbd_process () from /usr/sbin/smbd
#9 0x00002b55a5f00d00 in main () from /usr/sbin/smbd
# strace -p 32717
fcntl(10, F_SETLKW, {type=F_WRLCK, whence=SEEK_SET, start=15032, len=1}
From a netstat these processes are in a CLOSE_WAIT or a TIME_WAIT status.
Bye,
Marco.
--
To unsubscribe from this list go to the following URL and read the
instructions: https://lists.samba.org/mailman/listinfo/samba
