PANIC: open_mode_check:/CPU hog: still an issue in 2.2.5 on solaris?
jeff.mandel at probes.com
Wed Aug 28 22:30:59 GMT 2002
I've been troubleshooting a cpu race condition that looks like a problem
that's been around for a while on solaris.
I upgraded to 2.2.5 and was unable to get any handle on what could cause
this, I downgraded to 2.2.2 which was the last one I could remember
working well for us and got this before too long:
[2002/08/29 08:46:59, 0] smbd/open.c:open_mode_check(555)
open_mode_check: exlusive oplock left by process 4768 after break !
For file netscape/cookies.txt, dev = 3052111, inode = 1935369. Deleting
it to continue...
[2002/08/29 08:46:59, 0] lib/util.c:smb_panic(1055)
PANIC: open_mode_check: Existant process 4768 left active oplock.
[2002/08/29 08:49:05, 0] smbd/oplock.c:request_oplock_break(1026)
request_oplock_break: no response received to oplock break request to
pid 4768 on port 24655 for dev = 3052111, inode = 1935369
for dev = 3052111, inode = 1935369, tv_sec = 3d6dc2a7, tv_usec = 88146
The related smbd rose to 50% cpu usage and had to be killed with a kill
-9. I've set oplocks = no now to see if that will help, though this has
also happened in 2.2.5 with oplocks off.
This problem occurs in 2.2.5 where smbds just grow in cpu usage without
apparently doing anything, I just don't get the PANIC: open_mode_check
error in the logs.
It seems fairly well distributed amongst the various clients, though
some seem to start this more than others. There are only a few clients
(15) sharing on a server with mail and oracle. There's 4G of memory and
2x750Mhz cpu. I'd say is a very lightly loaded system.
I've seen some of Fredrik Ohrn's postings with truss and lsof, and it's
not clear that it's ever really been fixed.
What can I do to narrow in on the cause of this? Am I trading different
problems here between 2.2.2 and 2.2.5 or are these cpu races recognized
as the same problem?
More information about the samba-technical