Looping ruby and dispatch processes on apache

Hi !

I have ruby 1.8.6, rails 2.1.0 in Apache 2.0.59 / mod_fastcgi 2.4.6 on
RedHat Enterprise.

The first 2, 3 requests are served nicely and then latency for a page
load jumps to 40 seconds and more. On the server I get multiple
dispatch.fcgi processes looping forever and driving CPU load to 100%.
The page I look at is a simple hello world...

So I deinstalled fastcgi, installed the newest mod_rails / passenger
and hello world worked. Before porting the complete application I
build a login-page and after some twisting I get the pages running
nicely. After two tests page latency is going up, 40 seconds and more
and on the server I get, guess what, looping ruby interpreter
processes, driving performance of the CPU to 100%.

The same pages,hello world, login and the complete application,
however, render nicely in a local mongrel server in eclipse wuth 1.8.6
ruby, 2.1.0 rails on a Windows XP machine.

What is going wrong on my server ? I am getting desperate a bit by
now...

Hi !

I have ruby 1.8.6, rails 2.1.0 in Apache 2.0.59 / mod_fastcgi 2.4.6 on
RedHat Enterprise.

Might you be using one of the 'bad' releases of ruby ? ruby 1.8.6 p230
is known to segfault, leak memory etc...

Fred

Might you be using one of the 'bad' releases of ruby ? ruby 1.8.6 p230
is known to segfault, leak memory etc...

Don't think so.

I have the same version locally on the client / development desktop
and on the server. On the development desktop everything is just fine.
NO problem at all. On the server, the only problems I get is the
described one. The first two, three pages are always fine.

Might you be using one of the 'bad' releases of ruby ? ruby 1.8.6
p230
is known to segfault, leak memory etc...

Don't think so.

It's pretty easy to be sure: just run ruby -v and see what it says

Blushing red...I don'tier believe it...it really *does* say patchlevel
230, and I already started installing a new Apache 2.2 release...

I'll check if the problem is solved by moving to a different ruby
release and will report the results.

But: If this release is so bad, why is it still the one which is on
download at rubyonrails.org :-0

Might you be using one of the 'bad' releases of ruby ? ruby 1.8.6
p230
is known to segfault, leak memory etc...

Don't think so.

It's pretty easy to be sure: just run ruby -v and see what it says

Blushing red...I don'tier believe it...it really *does* say patchlevel
230, and I already started installing a new Apache 2.2 release...

I'll check if the problem is solved by moving to a different ruby
release and will report the results.

But: If this release is so bad, why is it still the one which is on
download at rubyonrails.org :-0

not rails' fault, they don't control what the ruby core team do (but
yes, the situation with ruby 1.8.6p230 is a mess. a newer release
without the segfaulting was released a day or two ago)

Fred

> I'll check if the problem is solved by moving to a different ruby
> release and will report the results.

Thanx Fred, for your continued support. Now it is:

Apache 2.2.9, installed from scratch
Passenger 2.0.3, installed from scratch
Ruby 1.8.7 patchlevel 71, installed from scratch

and it is running just nicely. Guess it was the ruby release.

I agree, it is not rails' fault if the ruby interpreter breaks. But
why does the rails team have a broken ruby on their (ie. rails)
download page. The broken ruby did not even produce a simple hello
world correctly in the described config - so it should have been found
in one of the simpler regression tests.

THANX a lot for your help, I already was getting a bit desperate.

Case closed !

I'll check if the problem is solved by moving to a different ruby
release and will report the results.

Thanx Fred, for your continued support. Now it is:

Apache 2.2.9, installed from scratch
Passenger 2.0.3, installed from scratch
Ruby 1.8.7 patchlevel 71, installed from scratch

and it is running just nicely. Guess it was the ruby release.

I agree, it is not rails' fault if the ruby interpreter breaks. But
why does the rails team have a broken ruby on their (ie. rails)
download page. The broken ruby did not even produce a simple hello
world correctly in the described config - so it should have been found
in one of the simpler regression tests.

It was a bit of a sticky situation: the previous version had some
security holes that had just been found, the fixed version segfaults.
The non segfaulting version has only just been released, i'm sure the
link will be updated shortly.

Fred