[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Watch your LDAP locks



On a general note, it would be good for anyone running ZCS to familiarize themselves with:

<http://wiki.zimbra.com/index.php?title=Performance_Tuning_Guidelines_for_Large_Deployments#Zimbra_OpenLDAP_Server>

and

<http://wiki.zimbra.com/index.php?title=Performance_Tuning_Guidelines_for_Large_Deployments#Configuring_the_BDB_subsystem_to_increase_LDAP_server_performance>

the latter of which discusses lock/lockers/lock objects

Regards,
Quanah


--On Monday, August 04, 2008 11:00 AM -0500 Tom Golson <tgolson@tamu.edu> wrote:

Hi Steve,
	I'm really surprised by this.  Wow.  We're not using distribution lists
(yet), but we have a fairly large directory which is being updated pretty
regularly.  We haven't come close to the limits, yet, for concurrent
locks or lockers.  What do your cache settings in DB_CONFIG look like?
Thanks!

--Tom

Steve Hillman wrote:
Hi folks,
  This list has been pretty quiet lately, so I thought I'd share a
  little nugget of info we came upon this week.

If you have a fairly large LDAP directory, or large objects (e.g.
distribution lists with lots of members), you may encounter lock issues
with Zimbra's LDAP directory. The default settings allow for 3000 locks
and 1500 locked objects. While I'm not clear on exactly how this relates
to the size of the objects in the directory, it's something worth
watching if you're of any size.

We encountered LDAP lock errors this week when trying to restore a
user's data. It was failing to add the user to our "all-employees"
distribution list. You can find out how many locks you're allowed, and
how many is the maximum your LDAP server has ever used, by becoming the
zimbra user, then "cd /opt/zimbra/openldap-data" and "db_stat -c". Ours
showed that we'd hit the limit at least once before (compare the
"maximum locks" numbers to the "maximum locks..possible" numbers). I
modified DB_CONFIG (in that same directory) to triple the limits then
restarted ldap. A few minutes later I ran the db_stat command again and
we'd already used more locks than the previous limit had been set to, so
we're not sure what other things may have been failing in the past.  In
any case, once the lock limit was raised, our restore succeeded.




--

Quanah Gibson-Mount
Principal Software Engineer
Zimbra, Inc
--------------------
Zimbra ::  the leader in open source messaging and collaboration