httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dimitri Rebrikov <dimitri.rebri...@t-systems.com>
Subject ap_get_client_block blocks in Apache 1.3
Date Wed, 09 Apr 2003 08:59:04 GMT
Hi!

We have problems with broken POST-Requests that our Partner sends to us.
They are using the Software wrote by they own.(Not a Web-Browser).
We are using Apache, mod_ssl/openssl, mod_jk, Tomcat und servlets to Process they
Requests. The POST-Requests they sends have from time to time too big Content-Length
declaration that the Requests' Bodies real have.
Such requests hangs as long as the Partner shut it down (long time!).
The httpd.conf Timeout parameter seems to dont play any role here.
Many such requests from many partner's instanzes paralyze our System.

Our enviroment: Apache/1.3.27

We have tested on Linux/Sun with/without HTTPS with the same results.

I have analyzed the Sorce Code from mod_jk/Apache and log-Outputs with
following cognition:

It hangs as the mod_jk try to read out the Body of the Request
using "ap_get_client_block" APACHE-API-Function.

I have readed the "Reading Client Input in Apache" - Tutorial
(http://httpd.apache.org/docs/misc/client_block_api.html) and can
say that the way the mod_jk do it is 100% conform to this.

I have tried mod_cgi on the Place of mod_jk and found out that mod_cgi would
also block without ap_hard_timeout set just before ap_get_client_block.
(I can see it in the apache-error-log). In "Reading Client Input in Apache" - Tutorial
is no mention about ap_hard_timeout before ap_get_client_block.

After all that i have following question:

Is it the correct behavior that ap_get_client_block hangs in our case?
(this means a Bug in mod_jk. The mod_jk must set ap_hard_timeout before ap_get_client_block.)
or is there a Bug in Apache 1.3?

Thanks in advance
Dimitri



Mime
View raw message