Mailing-List: contact tomcat-dev-help@jakarta.apache.org; run by ezmlm
Message-ID: <3F714C34.8050806@apache.org>
Date: Wed, 24 Sep 2003 09:48:04 +0200
From: Henri Gomez <hgomez@apache.org>
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; fr-FR;
 rv:1.4) Gecko/20030624
MIME-Version: 1.0
To: Tomcat Developers List <tomcat-dev@jakarta.apache.org>
Subject: JK add-ons to fix problems with HUNG tomcats...
References: <Pine.GSO.4.21.0309231454090.19227-100000@hume.nlm.nih.gov>
 <bkr03n$rfo$1@sea.gmane.org>
In-Reply-To: <bkr03n$rfo$1@sea.gmane.org>
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Transfer-Encoding: 7bit

Hi to all,

After the yesterday post of David Rees (mod_jk does not detect a hung 
Tomcat), I take a look at ajp/jk and make some works to fix the problem.

In fact the problem is in jk since the begining as jk use sockets in 
blocking mode.

The works I conducted :

- Modify ajp protocol on java side to accept a PING message (0x08).
   This message will be answered by low level ajp layer by a ..
   PONG(0x09).

- On the jk (not jk2 for now) I added 3 optional timeouts :

   a connect_timeout, which when set will send a PING request
   after connection and wait for PONG reply up to connect_timeout ms.

   a prepost_timeout, which when set will send a PING request before
   trying to send the real request to tomcat.

   in fine I added the long awaited reply_timeout, which will make
   jk wait up to reply_timeout ms before failing in error.


With such 3 timeouts (optional in workers.properties), we should be
able to detect any problem with remote tomcats and as such make a
better use of tomcat clusters via Load-Balancing.

If nobody object, I'll commit it this afternoon CET for review and
comments.

Note, that by default jk works as before....