Hi there, I would like to explain my scenario, perhaps this has been answered on this forum.
A bunch of random Ajax requests from the browser (IE9) end up with a 400 error code on the apache webserver and the the browser hangs for 5 minutes. Httpwatch shows the error code as ERROR_INTERNET_CONNECTION_RESET and then immediately afterwards IE fires the same request again, which shows up with a time taken of 5 minutes and error code as ERROR_HTTP_INVALID_SERVER_RESPONSE. The browser recovers after 5 minutes. Further investigation on the webserver and appserver logs reveals the following:: a.. The request comes from the browser and hits the webserver and then forwards to the appserver instantly. b.. The mod_jk log for the request shows that there is time duration of 5 minutes spent in the ajp_read_fully_server::jk_ajp_common.c(1399): enter. After 5 minutes I get the next line as follows ajp_read_fully_server::jk_ajp_common.c(1432): exit. Then in the next line i see the following ajp_send_request::jk_ajp_common.c(1766) worker 11 browser stop sending data, no need to recover. Later it shows unrecoverable 400, request failed. c.. The forensic.log show the content length as a nonzero value. d.. The applcation server log hangs in the org.apache.coyote.ajp.AjpProcessor.read method for 5 mintues and the continues the execution. The thread dump also confirms this. The questions I have are:: a.. Is this a problem with IE only because of the keepalive timeout and the apache webserver keepalive time(current value is set to 5seconds) out which is not in sync. b.. Is this a problem with the appserver not able to process requests that are bad/incomplete. c.. Should I increase the Apache webserver timeout value to 60s or more , will this have any performance impact. Kindly advise on the scenario. Many thanks for reading through. Current setup: Apache 2.2.24 Mod_jk 1.2.37 Redhat Linux VM JBoss EAP 6.1.0 JSF 2.1, Richfaces 3.3.4 Warm Regards Razi A. Ansari
