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

slapd slow


I have P4 2.4G , with linux softraid raid5 (with ide disks) and 1G ddram.

2 openldap 2.0.23 ( debian woody) servers are running there.
One is master, other is replica.

who use ldap: qmail-ldap courier-imap and and one linux server with ~20
active users logged in (pam-ldap nss-ldap  and samba 3.0.7 which of course
uses ldap).

index conf in both slapd.conf -s:

index   objectClass                             eq
index uid,mail,cn,sn,memberuid pres,eq,sub

Well thing is that quite often slapd is slow, machine load is quite high
(caused by slapd).

So far it's been not very critical.

But I tried to move my qmail+courier system (qmail and courier both use
ldap "directly", without pam-ldap or nss-ldap) to
postfix+courier+nss-ldap+pam-ldap architecture (where only postfix uses
ldap directly for lookups, other apps use nss and pam).
First I thought that nscd cacheing makes things faster, but then I
discovered that
1) pam-ldap has no cacheing
2) postfix+pam-ldap generate more ldap queries per operation than it was
with qmail-ldap+courier-ldap
3) and couple of test emails coused immediately high load, slapd being on
top of process list.

Anyway, hardware seems to be good enough, is there something important
that im missing in slapd conf?
Or is slapd supposed to be so slow , and I supposed to get better hardware?

Would upgradeing to openldap 2.2.x make any difference in performance?