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

Re: Slapd crashes often on Solaris 8



I've had quite a lot of bitter experience with this. The received wisdom seems to be that it's an OS issue. It appears that there's a memory leak in recomp() in libc. Since this is kinda fundamental I would doubt if cache / threading makes much difference. Sun say there's no patch available at the moment but says it will be fixed in the next minor release.

On a more positive point, I soak tested 2.0.7 on solaris 2.6 over the weekend (mulithreaded with TCP wrappers) - 3.2 million queries in approx 60 hours. Memory useage was approx 10 MB with only 300 KB added since slapd was started.

Perhaps I could suggest other sufferers hassle Sun Microsystems ??!!

-ian

----------------------------------------------
Ian C. Smith,
e-mail: i.c.smith@liverpool.ac.uk
Address: Computing Sevices Department,
        Brownlow Hill Building - room G11,
        The University of Liverpool,
        Liverpool L69 3BX.

--On Wednesday, February 07, 2001 8:19 AM -0800 "Kurt D. Zeilenga" <Kurt@OpenLDAP.org> wrote:

At 11:10 AM 2/7/01 -0500, Santithorn Bunchua wrote:
I have seen a few posting about Slapd stability
under Solaris 8 but haven't found any solution.

experiment with cache settings try using a different regex library try without threads

Is there any solution yet? The size of slapd
process still grow and grow fast and from
times to times used up all our gigabytes of
swap space. Now I have to restart slapd
three times a day by cron. I am using
slapd 2.0.7.

Thanks,

keh