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

Re: 2.2 bdb leaks?





--On Wednesday, December 08, 2004 2:08 PM +0900 Carsten Haitzler <raster@valinux.co.jp> wrote:

ok... someone please tell me i'm wrong.

i have a 500,000 entry (inetorgperson) db - sot its big enough (for
testing)... and i'm playing with improving bdb/hdb... and i seem to find
it seems to allocate ram... till the cows come home.

basicalyl i have set the cachesize to 10 (and idlcachesize to 8)... i keep
querying for items from a client (going thru uid 0 and up numerically)
and slapd just keeps allocating ram... and allocating.. and allocating.
(dmalloc, valgrind, memprof side checks)... this smells wrong to me. if i
keep querying eventually i get slapd to be using > 1gb of ram...

something smells wrong - just to save me hunting for something i should
be.. is this right? why? shouldn't it level out at keeping 100 entries in
the cache and thats it? that means memory use should have leveled out
after the cache fills?

(hunting for more "culprits" that could be the issue) just wondering if
i'm barking up the wrong tree here.?

What version of 2.2 are you using, that would help...

Note that 2.2.19 had some memory leak & cache leak fixes.

--Quanah

--
Quanah Gibson-Mount
Principal Software Developer
ITSS/Shared Services
Stanford University
GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html