Red Hat Bugzilla – Bug 69794
OpenLDAP 2.1 not as stable as 2.0
Last modified: 2015-01-07 18:57:56 EST
From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.0.1) Gecko/20020712
Description of problem:
Sorry that this may not be an extremely helpful report, but... I migrated my
schema and configs from the OpenLDAP 2.0.23 to the OpenLDAP 2.1.2/2.1.3 in
Limbo/Rawhide and although performance increased significantly, it became rather
unstable under load.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Clear LDAP database
2. Start an import 6000 LDAP entries (via PHP)
Actual Results: Somewhere around 5000 entries, LDAP will fail with either an
unknown error or a "Critical Component Unavailable" error.
Expected Results: Nothing... it should just keep chugging like 2.0.23 does.
The data set being imported is identical to the set being imported into
2.0.23... the only difference is the version, and therefore the bdb backend. My
suggestion to RedHat is to write a script that will similarly make thousands of
reads/writes to LDAP and see if it is reproducable in your environment.
I did not try 2.1 with any of the other backends. I suspect they may be more
reliable, though. I also noticed that upon restart of the server after the
crash, it took about 30 seconds before the server would respond to queries.
Again, this makes me think that something is amiss with the backend... like it's
reindexing upon restart. These are all guesses, though.
I'll be happy to help out with testing... It took until 7.3 for the 2.0 builds
in the Redhat 7 series to become stable (at least for us on about 6 different
boxes we ran at various times), so I'd hate to see a step backwards in
reliability by upgrading to 2.1 prematurely.
We'll look at 2.1 for a future release.