Re: slapd eats up memory- database corruption ??

["Dejan Muhamedagic" <dejan.muhamedagic@at.ibm.com>]

Hi,

The dbcachesize option is said to be one per index, which means that if 
you have more than one index, which you probably do, then you can't obtain 
that much memory per index (.5gb).  Well, the system (wherever is the 
culprit) should tell so, and maybe it does somewhere, but that's another 
story.

Cheers.

Dejan
Please respond to Susanth_TS@sifycorp.com 
Sent by:        owner-openldap-software@OpenLDAP.org
To:     openldap-software@OpenLDAP.org
cc: 
Subject:        slapd eats up memory- database corruption ??



                                              Placed At :


Hi All,

I have been running slapd for the past 2 months on openldap version 
2.0.17.  The
problem is face all of a sudden is that slapd eats up the full memory and 
dumps
core( it works fine for an hour or so and then the memory gets eaten up 
slowly
and dumps core). I upgraded the version to 2.0.23 and found that the 
memory gets
eaten up and then not able access the machine at all.

When the problem occurs i have noticed from the log that reads happen from 
ldap
and writes are not happening.

1. Is it because of some corruption in the berkely DB database ?

2. how do i find if there is corruption in  backend database and resolve 
it ??

3.  Is there some memory leak in slapd ?

4. Have set the following parameters

     cachesize 10000
     dbcachesize 500000000

The machine has 1 GB of RAM . What is the value that need to be set for 
the
above parameters has it goy anything to do with writes on to the database

This is very urgent.. please can someone respond

Waiting for the reply.. and thanks in advance