Oh, hi again :)!
On Tuesday 23 November 2010 07:15:27 pm Sebastien Caty wrote:
> I've hit a problem on some servers that are under heavier samba use. I
> have no proof at all so far that SVS in causing this but I'm
> investigating. Since logging level was fairly low, I don't know how
> this happened but the end result is a directory gets locked. A user
> cannot open this directory (explorer freeze). The user close explorer
> and tries to open it again on the same directory. Each time a user
> does this on the locked directory, another smbd process is spawned and
> stalled in fnctl F_SETLKW. So it's waiting forever for a lock to be
> released. Only way to make it work again is to kill all stalled smbd
> process.
I've seen this before... but not since 0.1.1 was released.
> This is the first time I've seen this happen since upgrading from
> samba 3.0 to samba 3.5.6 and SVS. I don't know how to reproduce this
> yet. Right now I have disabled SVS and waiting a week see if it
> happens again. If not, I'll enable SVS again but with more debug
> logging.
No, it's probably really caused by SVS. Samba will just respawn a new session
(and yet another new session)... which is why you may end up in a number of
stalled SMBD processes.
But that's hard to reproduce, I know... well, if it's still there, I have to
try to find it.
> I don't know why I end up with many machine connection for the same
> user.
You only have one active connection, but the SMBD processes hang.
> I've also enabled reset on zero VC to make samba close all
> previous process from the same IP. But even with the reset, I've seen
> samba get stuck with many smbd all waiting for a lock.
>
> Might be some wierd race condition/deadlock.
Yeah, most probably a deadlock. Give me time to study it... also, please let
me know your svs.ini.
I may not find the time for it this week, but I'll do it ASAP.
Thanks, René
--
René Reucher
rene.reucher@...
http://www.batcom-it.net/
Health is merely the slowest possible rate at which one can die.
|