Greetings, oh ye Mod_Clusters masters!
Our 'mod_cluster 1.1.0-Final' rig is configured to support SSO by way of SiteMinder, which has been working very well. We recently onboarded an application which started exhibiting errors like the following ...
[Fri Jan 13 22:08:21 2012] [error] ajp_msg_append_cvt_string(): BufferOverflowException 4 7524
[Fri Jan 13 22:08:21 2012] [error] ajp_marshal_into_msgb: Error appending the header value
[Fri Jan 13 22:08:21 2012] [error] ajp_send_header: ajp_marshal_into_msgb failed
[Fri Jan 13 22:08:21 2012] [error] (120001)APR does not understand this error code: proxy: AJP: request failed to aaa.bb.c.ddd:eeeee (aaa.bb.c.ddd)
This symptom prevented application requests from completing, but What was ununsual was that only a subset of users were affected. We soon found out that the affected users' requests carried an unusually large payload (>8KB) due to large header values that were being inject by SiteMinder. Switching to HTTP-connector yielded a similar error, so it's not necessarily an AJP fault. We've seen positive results when increasing request buffers sizes on HTTPD and Tomcat sides, but I'd like to understand the nature of this symptom.
We found some useful information on the AJP front here ..
... but is this a general request processing bug in between HTTPD and Tomcat regardless of protocol for which the only answer is to increase buffer sizes?