Description of problem:
Satellite 5.2 on x86_64 platform reports the following error in /var/log/httpd/error_log:
Apache2::SizeLimit httpd process too big, exiting at SIZE=400136 KB SHARE=12560 KB REQUESTS=153 LIFETIM
E=4013 seconds
The message comes every other second when deploying a new machine. This means that the deployment of servers are extremely slow to the point where the Satellite becomes unusable.
How reproducible:
Install Satellite 5.2 on x86_64 HW and deploy a server.
Steps to Reproduce:
Actual results:
It appears that each time a satellite is installed on a x86_64 box, the default MAX_PROCESS_SIZE is too low and has to be increased in order to avoid apache being killed because it eat too much memory.
Deploying a server takes 30 minutes instead of five. The server load is heavy because httpd-processes are killed, a new is forked which grows to the SizeLimit and then gets killed again.
Sample errors:
[Fri Dec 26 10:01:06 2008] (2022) Apache2::SizeLimit httpd process too big,
exiting at SIZE=526616 KB SHARE=13580 KB REQUESTS=2 LIFETIME=0 seconds
[Fri Dec 26 10:01:08 2008] (2034) Apache2::SizeLimit httpd process too big,
exiting at SIZE=526708 KB SHARE=13548 KB REQUESTS=2 LIFETIME=0 seconds
[Fri Dec 26 10:01:13 2008] (2044) Apache2::SizeLimit httpd process too big,
exiting at SIZE=526708 KB SHARE=13548 KB REQUESTS=2 LIFETIME=0 seconds
...
Expected results:
Deployment of server should be a lot faster and httpd-process should not be killed.
Temporary workaround:
Edit /etc/rhn/satellite-httpd/conf/startup.pl and changing the
MAX_PROCESS_SIZE to a size larger the the processes are reported to getting killed at (ideally stay multiples of 8), after that everything works much quicker and without errors.
Satellite team:
What happens if you remove Apache2::SizeLimit, and rely on the OOM killer? Does it still die?
We had a very similar problem on the www-perl stack, it turned out that Apache2::Sizelimit was measuring memory usage in a way that wasn't quite right on 64 bit boxes, and it ends up killing Apache when it shouldn't be killed. We bumped MAX_PROCESS_SIZE up as a workaround, but we'll likely be removing SizeLimit in the next release.
An advisory has been issued which should help the problem
described in this bug report. This report is therefore being
closed with a resolution of ERRATA. For more information
on therefore solution and/or where to find the updated files,
please follow the link below. You may reopen this bug report
if the solution does not work for you.
http://rhn.redhat.com/errata/RHEA-2009-1434.html