[Spacewalk-list] spacewalk eats up all connections to postgres

[Ree, Jan-Albert van]

>

Jan-Albert van Ree
Linux System Administrator
MSuG MARIN Support Group
E mailto:J.A.v.Ree at marin.nl
T +31 317 49 35 48M +31652598885

MARIN
2, Haagsteeg, P.O. Box 28, 6700 AA Wageningen, The Netherlands
T +31 317 49 39 11, F , I www.marin.nl

-----Original Message-----


> On 12/7/2012 09:59, Ree, Jan-Albert van wrote:
> > I've applied the patch yesterday as well, and indeed this does seem
> to help a lot.
> > I'm only seeing about 35 connections now (with 25 clients), no more
> hanging queries either.

> I too applied the patch this past Saturday and disabled my nightly
> application restart cron to see how things managed, and the patch seems
> to have not only fixed my issue with idle connections (70 clients, 5
> total "idle in transaction" processes) but also cleared up some issues
> I was having with spacewalk's internal scheduled tasks. I have not
> needed to restart my Spacewalk 1.7 (on CentOS 6.3) since.
> I'm in the process of migrating to a fresh new 1.8 install, but I'll
> slow that down a little bit as this "idle transaction" issue was the
> driving force behind the migration. I'll update the list after a good 5
> "production" days.

It's been almost a week now and connections are around 50, with no 'idle in transaction'
To sum it up, my fixes were

- Update the postgresql jar under /usr/share/java to latest version from jdbc.postgresql.org
- I changed TCP/IP parameters on PostgreSQL and Spacewalk machines : net.ipv4.tcp_keepalive_time = 600
- Updated python-psycopg2

I now have Spacewalk 1.8 running stable with OSAD enabled, on PostgreSQL 8.4.13 (stock ScientificLinux RPMs with minimal tuning) I'm not sure if all the above steps are actually needed but I'm listing them anyway.
--
Jan-Albert van Ree