Mailing-List: contact tomcat-dev-help@jakarta.apache.org; run by ezmlm
Precedence: bulk
Reply-To: "Tomcat Developers List" <tomcat-dev@jakarta.apache.org>
Message-ID: <018901c22cb7$3096c620$6501a8c0@kasim>
From: "Arshad Mahmood" <arshadm@compuvision.co.uk>
To: "Tomcat Developers List" <tomcat-dev@jakarta.apache.org>
References: <20020715155602.L63610-100000@icarus.apache.org>
Subject: Some advice regarding mod_jk
Date: Tue, 16 Jul 2002 11:54:12 +0100
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit

Hi,

I am developing an using mod_jk, apache 1.3.26 and Tomcat 4.1.7, Redhat 7.2
and have come across an issue with regards to mod_jk.

Now I assumed that the Ajp13Processors under mod_jk acted as pool to process
incoming requests. But I hadn't realised that a processor is created per
socket and then it processes requests on that socket until it is shut down.

This means that even on a lightly loaded machine I can quite easily have 7-8
processors (hence threads, hence processes) depending on which apache child
processes my request.

I would like to modify the code for the Ajp13Connector as follows, and I
wonder what the developers think the consequence would be for reasonble
sized sight (say ~10,000 hits a day).

1. Add a maxIdleProcessors, so that short periods of peak activity don't
leave loads of processors hanging around. I would probably set this value
quite low (probably 2-3).

2. I would modify Ajp13Processor so that it processes a single request and
then recycles itself, rather than waiting for more requests on the same
socket.

I have made these changes and tested them and the behaviour doesn't appear
to be too bad. Peak activity does slow a little bit, but overall usage of
resources (i.e. threads) is much lower.

Any comments?

Regards,
Arshad


--
To unsubscribe, e-mail:   <mailto:tomcat-dev-unsubscribe@jakarta.apache.org>
For additional commands, e-mail: <mailto:tomcat-dev-help@jakarta.apache.org>