[Date Prev][Date Next]
[Chronological]
[Thread]
[Top]
Ldap or bdb hangs....
Hi all,
The problem here is the openldap or bdb hangs for some reason, making that
slapd doesn't accept no more connections for a while and ldapsearch is very
slow... And this happen when the client establishes a few 20/30 connections
do the ldap server and should be handling with about 400/500 connections per
second...
Openldap is installed in a server with dual XEON 2.8Ghz with 2GB of memory,
and has a DB with about 20000 entries to authenticate PAM_LDAP from a
solaris (migrated from NIS+), meaning that I'm using only posixAccount,
shadowAccount and account object classes.
The openldap version is 2.2.17 and BDB is 4.2.52.2!
I've read this article http://www.openldap.org/faq/data/cache/1075.html and
followed this thread
http://www.openldap.org/cgi-bin/wilma_hiliter/openldap-software/200406/msg00
258.html?line=3#hilite
And that's why my configurations file are as follows:
SLAPD.CONF
conn_max_pending 1000
idletimeout 15
sizelimit 25000
threads 2000
timelimit 30
#######################################################################
# BDB database definitions
#######################################################################
database bdb
cachesize 150000
directory /db
checkpoint 1024 30
# Indices to maintain
index default pres,eq
index objectclass
index uid
index uidNumber
access to * by * read
*****************************
And here you have DB_CONFIG file:
set_lg_bsize 262144
set_lg_max 2097152
set_cachesize 0 536870912 1
***********************
I've made some calculations from the article mentioned above and 4096*(7+1)
+ 16384 * (1+1) = 65536 bytes, meaning that this is the only cache that I
need.... But I've to mention that before these kind of "tuning" was maded
the behavior was excatly the same
I think the reason of this bad performance have to reside in some
configuration issue and I hope someone could help me with this...
TIA
Hugo Tavares