tomcat-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jim Jagielski <...@jaguNET.com>
Subject Re: possible mod_jk "feature"
Date Fri, 18 Aug 2006 20:36:25 GMT

On Aug 18, 2006, at 4:06 PM, Mladen Turk wrote:

> Jim Jagielski wrote:
>> On Aug 18, 2006, at 12:40 PM, Mladen Turk wrote:
>>> The Tomcat should be responsible for deciding
>>> if the connection will be reused or not.
>> This doesn't match, iirc, how the whole cache_timeout
>> (or connection_pool_timeout) is done within mod_jk...
>> Basically, what's needed is a way to make each
>> AJP a one-shot.
>
> Right, but it should be in the form:
> 1. mod_jk/mod_proxy_ajp connects
> 2. Tomcat responds if the connection
>    'should/should not' be closed after the
>    response.
>

My reading is that the reuse flag is simply tagged
to whether or not we rec'd anything (basically)
at the END_RESPONSE phase. Even if we do, we
still force a reuse (which is weird since we
log the fact that we're not going to reuse it).
We do set reuse to FALSE if we get a client error (and
we have an abort-on-error set). In fact, that's
how we mark it for closing. In any case, if reuse
is FALSE, ajp_reset_endpoint does the right thing.

So, IMO at JK_AJP13_END_RESPONSE, we should:

    1. Not reset reuse to TRUE on a protocol error
    2. Check for a 'r->never_reuse' flag and, if set,
       force reuse to FALSE.



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


Mime
View raw message