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

Calculate slapd memory footprint



OpenLDAP 2.1.25 using back-bdb
Berkeley DB 4.2.52 + 1 patch
RedHat Linux 9.0

Since our switch to back-bdb last weekend, I have noticed that both our
master and replica slapd processes have yet to stabilize at a particular
memory footprint.  We've restarted both servers once last Friday after
installing a kernel patch, so I don't know whether we just haven't waited
long enough for them to stabilize, but I thought I'd ask:

How would one calculate the expected stable memory footprint of slapd?

Our master slapd memory footprint is currently 123MB (about 12% of total
memory) for code plus data plus stack space (the 'SIZE' column in top)
and is still gradually increasing.  Is that a reasonable figure? It
seemed rather high to me given the following:

Our BDB cache size (calculated from page sizes, hash buckets, et al,
then increased to taste):

set_cachesize   0       11470528        1

Our entry cache size:

cachesize	1000

Total entries: about 8000

Number of attributes indexed: 22

Size of an LDIF backup: about 7.5MB

Though it may not be relevant, our BDB cache does seem to be properly
sized:

# slapd_db_stat -m | egrep "(Clean|pages found)" | head -2
166M    Requested pages found in the cache (100%).
26247   Clean pages forced from the cache.


-- 
Kirk Turner-Rustin
Information Systems
Ohio Wesleyan University
http://www.owu.edu
ktrustin@owu.edu