Mongrel_cluster : up to 500 MiB memory


#1

Hi everyone,

I’ve got some trouble with my RoR application running with apache
(2.2.9) and mongrel (1.1.5).

The site i manage has about 20’000 visits a day, and we run 14 mongrel
instances switched with mongrel_cluster (1.0.5). The fact is, sometime,
half of my mongrel instances are raising something like 500MiB memory
and don’t seem to work anymore.

The server itself has 8 CPU’s and 8Go of RAM.

I read different how-to’s about mongrel and mongrel_rails but most of
them are just talking about how to start with mongrel.

I would like to know if there is a way to monitor mongrel instance
and/or how to improve my configuration?

If anyone has a suggestion, it’ll be very helpfull.


#2

You didn’t specify exactly what you want to monitor but a lot of
people use Monit to monitor their mongrel clusters:
http://mmonit.com/monit/

I have also heard people talk about using ‘god’

http://weblog.redlinesoftware.com/2008/4/4/using-god-to-monitor-mongrels

Quoting C??d B. removed_email_address@domain.invalid:


#3

Céd B. wrote:

Hi everyone,

I’ve got some trouble with my RoR application running with apache
(2.2.9) and mongrel (1.1.5).

The site i manage has about 20’000 visits a day, and we run 14 mongrel
instances switched with mongrel_cluster (1.0.5). The fact is, sometime,
half of my mongrel instances are raising something like 500MiB memory
and don’t seem to work anymore.

There are some patches on ruby-core that discuss a sweller GC maybe that
would help.
http://www.ruby-forum.com/topic/170608#new