slapd process needs a restart

["Roberto Balducci" <roberto.balducci@proconsul-it.it>]

Hi, 

we're having a problem with our openldap installation.

Our software release are:

   - db-4.1.25.NC
   - openldap-2.1.23

and are installed on an alpha ds10 with TRU64 5.1A operating system.

Our db has about 4300000 entries and the id2entry.bdb file is about 7,7 gb.

We had some performance problem, so we had installed another 1 gb of ram, and better tuned the slapd, updating the slapd.conf and DB_CONFIG file: we set cachesize slapd.conf's attribute to 3000000 and the set_cachesize DB_CONFIG's attribute to 1 52428800 1; we had also updated the kernel related attributes.

Now slapd is faster and it has performance as we need, but after 3 days of work, even if it's running and seems to be working fine, it doesn't accept any new connections or become very slowly, and we do not have any other things that to restart it.

>From system point of view, nothing seems strange, socket are all established, but running the netstat -an | grep 389 command more times, we see that the number of bytes in either receive and send queue, do not change.

In this scenario, if we start ldapsearch, it takes 15 minutes, to retrieve information.

The ordinary work is to answer to thousands ldap query request per day and to add about 15000 entries per day.

Any suggestion will be very appreciate.

Ciao, Roberto