Description of problem:

The directory server processes on IPA servers don't seem to close file handles
properly so given enough time they run into the nofiles limit of 8192 at which
point they become unresponsive until the process is restarted.

The deleted (but not closed) files are named as /var/tmp/ldap_49X

Version-Release number of selected component (if applicable):
389-ds-base-1.2.9.14-1.el6_2.2.x86_64
389-ds-base-libs-1.2.9.14-1.el6_2.2.x86_64

How reproducible:

Always

Steps to Reproduce:
1. start IPA
2. run for x in `seq 1 100000` ;do ipa user-find > /dev/null 2>&1 ; done
3. ls -l /proc/<pidof ds slapd>/fd

Actual results:

The number of open file handles (deleted files) keep on increasing and slapd
stop responding when it reaches the max open file limit.

Expected results:

DS closes the file handles properly

Additional info:

I am unable to re-produce the issue completely, ie; on my test machines (with
less activities) the number of open file handles (of deleted file) are less
than 20, but we could create ~20 open file handles (of deleted files) by simply
executing a ipa user-find command in loop.