Jeremy Allison wrote:
On Tue, Jun 13, 2006 at 11:37:05AM -0400, Ryan Steele wrote:
I desperately need a resolution to this issue. I've asked once (about a
day or two ago), but I haven't heard anything back. The only reason I
press the issue is I may because without a quick resolution, I may be
forced to switch over to AD (cry!). I submitted a request via Bugzilla
but I saw a slightly similar problem with 3.0.20a that still hasn't been
resolved, so I thought this might be a quicker route? Here's a synopsis:
If you *have* to have a resolution on issues then you need to
buy support from someone, or use a Linux distro that has support
available and buy that.
Well, my life doesn't depend on it, I really just want to support Samba
and stick with it because I've never had problems before this and in
general I really like the Samba suite.
I navigate through Windows Explorer to My Network Places and so on until
I get to the server. I open up a folder on the server. I execute the
'ps auxwww | grep mbd' and sure enough, there's the share. The
smbstatus command confirms this. Now, I close out that Windows Explorer
window I have open to the server. However, a 'ps auxwww | grep mbd'
shows that there is still a connection open to this folder.... an
smbstatus confirms.... After a few minutes, the user for that pid
changes to root, and the process just sits in there forever, sucking up
0.9% of memory. This happens with EVERY share Windows opens, and when
it gets in this state, I can't open any new shares. I've tried using
the "deadtime" option to kill these...no dice, they still hang around.
In fact, the only thing that gets rid of them is a 'killall -9 smbd'.
There isn't a generic problem with Samba and XP, millions of clients
are using it successfully (and I'm not saying that lightly). If the process
seems stuck try attaching to it with gdb or strace and find out what
it's doing.
I'm glad to hear there isn't a generic problem. As far as strace, it's
a bit difficult in this situation; by the time I'd have the information
I'd need, there would be tens of thousands of lines of superfluous
information....making it near impossible to extrapolate the useful
stuff. It'd be like finding a needle in a haystack. Also, I'm a
competent coder, but not a great one...so stepping through it with gdb
might not be the best solution for me. I was hoping there might be a
way to attack this problem without having to step through it with, say, gdb.
Don't use kill -9, that can damage internal Samba databases.
Is there a better way? `/etc/init.d/samba stop` and `kill /pid/` don't
work...the processes still hang around.
If you don't know how to do these things then either learn to do so,
or buy support from someone who can. Sorry to seem harsh but the
reality of such pleas is that it's easier to help someone who knows
how to help themself.
It does seem a bit harsh. I fully support Samba and the wonderful
effort and strides you guys are making...it really does show! However,
not all of the users/administrators are of equal experience, so we turn
here for advice. On your advice though, I'll figure out how to strace
one of the processes generated by an XP client and include that in a
follow-up to the mailing list, along with whatever I'm able to
extrapolate from the data. In my Bugzilla report, I included an excerpt
from the log file that was generated by attempting to access share
information from the server's command line at such a time when the Samba
server was in an overwhelmed state, along with info from stdout. The
Bugzilla tracking number is #3832.
Jeremy.
Any thoughts? Thanks!
Best Regards,
Ryan
--
To unsubscribe from this list go to the following URL and read the
instructions: https://lists.samba.org/mailman/listinfo/samba
