[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: slapd problems under stress


Mark Valence wrote:
> Stephen-
> >I can send you the test program (I don't think it is interesting enough
> >to send it to the list...)
> Can you send me a tarball with source and a Makefile?

the source is relatively short (only one file) so the tarball doesn't
make too much sense. I already tried to send it to you, but I got an
error message from the mailer. (There seems to be a DNS problem

<kurash@sassafras.com>: Name service error for domain sassafras.com:
Host not

Do you have another mail address? (otherwise I will try to send the file
to you from a different mail account)

> >That seems to improve things a lot! I couldn't crash the system with the
> >ldbm-backend anymore if I apply that patch. I even get answers to all
> >the requests :-). I'll try our backends on monday.
> Sounds to me like a deadlock problem, where the worker threads are
> not completing, and so you get more and more threads (960!) on your
> system until you crash.  Can you experiment with values of
> max_concurrency (the second param above)?  Interesting values would
> be 1, 10, 100, 500, and 1000.  Which values are OK, and which aren't?

It works fine with 1 and 10 concurrent threads. With 100 threads it
still works, but the timing seems a little irregular. With 500 threads I
get into trouble. I don't get many answers and the slapd crashes on

ldbm backend done syncing
====> cache_release_all
ldbm backend syncing
ldbm backend done syncing
====> cache_release_all
slapd shutdown: freeing system resources.
slap_sig_shutdown: signal 2
slap_sig_shutdown: signal 2
[... a lot of repetions removed]
slap_sig_shutdown: signal 2

(Speicherzugriffsfehler means access violation)

I guess the optimum must be somewhere around or above 10 (at least in
this environment).


Stephan Siano                           Mail:  Stephan.Siano@suse.de
SuSE Linux Solutions AG                 Phone: 06196 50951 31
Mergenthalerallee 45-47			Fax:   06196 409607
D-65760 Eschborn