[Spacewalk-list] Spacewalk 1.7 Postgres out of connections

[Scott Piazza]

I recently upgraded from Spacewalk 1.6 to 1.7.  In this version, I am
seeing an issue where Spacewalk works fine for several hours, but then
the database runs out of connections, and Spacewalk begins throwing
errors.  Similar Postgresql errors on other forums try to blame the
application for not closing connections correctly.  Older Spacewalk
posts online seem to suggest there's an issue with osa-dispatcher that
might be causing it.  I have disabled osa-dispatcher for now and am
waiting to see if that helps narrow down the source of the problem.
However, if anyone has additional suggestions to get this fixed, I would
appreciate hearing them.


Server: CentOS 6.2
Spacwalk: 1.7
SELinux: off
PostgreSQL: 8.4.9-1.el6_1.1
Postgres max connections: 500

/var/log/rhn/rhn_server_xmlrpc.log:
2012/05/10 12:17:50 -07:00 29438 10.0.11.2:
rhnSQL/driver_postgresql.check_connection('ERROR', "DATABASE CONNECTION
TO 'whmgmt' LOST", "Exception information: Database instance has no
attribute 'dbh'")

/var/log/rhn/osa-dispatcher.log:
2012/05/10 08:14:09 -07:00 10255 0.0.0.0:
rhnSQL/driver_postgresql.check_connection('ERROR', "DATABASE CONNECTION
TO 'whmgmt' LOST", 'Exception information: server closed the connection
unexpectedly\n\tThis probably means the server terminated abnormally\n
\tbefore or while processing the request.\n')

/var/lib/pgsql/data/pg_log/logfile.log:
2012-05-10 17:50:47 MST - whmgmtuser - [local]: FATAL:  sorry, too many
clients already
2012-05-10 17:50:49 MST - whmgmtuser - 127.0.0.1: LOG:  unexpected EOF
on client connection
2012-05-10 17:51:13 MST - whmgmtuser - [local]: LOG:  could not send
data to client: Broken pipe