RE: Problems in subjecting openldap to load test...

["oberwetter, josh" <joberwetter@grownetwork.com>]

This might be totally unrelated, but is it possible that there is a file
descriptor leak? 

When I tried to import a large LDIF without turning indexing off, I
observed error messages that the machine had run out of file
descriptors. Then slapd crashed.

-----Original Message-----
From: Sovan_Shatpathy@satyam-infoway.com
[mailto:Sovan_Shatpathy@satyam-infoway.com]
Sent: Thursday, August 02, 2001 10:49 AM
To: openldap-software@OpenLDAP.org
Subject: Problems in subjecting openldap to load test...




Hi,
      Here we  have been subjecting openldap to heavy loads(load
testing) in
order to test for the stability of the same when there are peak loads.(
ie a
very large number of users are trying to authenticate at the same time
with
multiple processes also running which need to connect to the ldap
server)

In order to achieve the same we have been running four processes which
are as
follows :

1. One process adding users to the ldbm database at the rate of 3600
user per
hr.
2. Second process changing password at the rate of 3600 users per hr.
3. A third process disabling users at the rate of 3600 users per hr.
4. A fourth process which extracts user information from the LDAP  at
random and
running continiously.

Now with the above mentioned four processes running in the back end we
have
tried to authenticate users . The problem we have noticed is that for
some time
with all the four processes running in the back end also the speed of
authentication has been  very good. But the processes  after running for
some
time , stops as the slapd hangs. Hence authentication also comes to a
abrupt
halt.

When we ran the debug level 256 it gave an error " TOO MANY FILES OPEN,
CANNOT
CONNECT TO LDAP SERVER"

I  am pasting the error below:

daemon: accept(6) failed errno=24 (Too many open files)
daemon: accept(6) failed errno=24 (Too many open files)
daemon: conn=1060 fd=15 connection from IP=127.0.0.1:62930
(IP=0.0.0.0:0)
accepted.
conn=1060 op=0 BIND dn="CN=MANAGER,O=xyz" method=128
ber_flush: 14 bytes to sd 15
conn=1060 op=0 RESULT tag=97 err=0 text=
request 1 done
==>meta_back_dobind: (anonymous) bind as "cn=Manager, o=xyz" failed with
error
"Can't contact LDAP server"
==>meta_back_dobind: (anonymous) bind as "cn=Manager, o=xyz" failed with
error
"Can't contact LDAP server"
==>meta_back_dobind: (anonymous) bind as "cn=Manager, o=xyz" failed with
error
"Can't contact LDAP server"
==>meta_back_dobind: (anonymous) bind as "cn=Manager, o=xyz" failed with
error
"Can't contact LDAP server"
ber_flush: 56 bytes to sd 59
ber_flush: 54 bytes to sd 61
ber_flush: 53 bytes to sd 63
conn=1060 op=1 SRCH base="OU=WEST,o=xyz" scope=2 filter="(uid=l79)"
conn=1058 op=1 SRCH base="OU=UNKNOWN,o=xyz" scope=2 filter="(uid=l79)"
ber_flush: 14 bytes to sd 15
conn=1060 op=1 SEARCH RESULT tag=101 err=0 text=
ber_flush: 14 bytes to sd 60
conn=1058 op=1 SEARCH RESULT tag=101 err=0 text=
request 2 done
=>meta_back_search [0] match="" err=""

Now once this happens the only option left is to restart  the slapd..
and  then
start the processes again .

Any suggestions as to how we can overcome the problem??

Thanks in advance ...

regards
Sovan

..

Satyam Infoway Ltd