[Spacewalk-list] Jabberd's session manager going loop

[Maxime VEROONE]

Hi,

I can't really tell what started it, but since a few weeks ago, my spacewalk 2.3's jabberd is running differently than before.
Spacewalk server is a CentOS 6.11, managing about ~350 systems ranging from centos 5-32 to centos 7-64.

Now, despite having cleanned /var/lib/jabberd/db/* several times and rcreating auth client-side by deleting /etc/sysconfig/rhn/osad-auth.conf, I still face this roblem :

Jabberd's "sm" process consumes 20% CPU all the time, I bet plenty of I/O but 99% hits disk cache, and generates >1GB of Berkeley DB recovery log per day.
A few month ago, we didn't even have to bother cleaning up db logs, and now it's twice a day (using recommended procedure from doc through db_archive -d)
>From strace logs, the process is perpetually re-reading the sm.db file, and then sends routes to router process who forward them to the c2s process

That frequency of "doing things" does not seem to correlate with how often osa-dispatcher sends pingsn yet most of the inter-process communication seems to be between osa-dispatcher and c2s...

Well, all of this puzzles me, as I have no deep knowledge of jabberd... did anyone ever meet this problem ?

Maxime Veroone
Junior Linux Sysadmin @ Capensis, France