We have a situation where a customer can’t send us mime documents over https. In fact we receive them but the customer never receives the http response back. By tracing the packages we can see that webMethods “cuts” the connection very quickly by sending an END-OF-DATA package followed by a RESET package. It never waits for an ACK package from the client. Both packages are sent from WM within 0.002 seconds. This triggers a broken pipe exception on the client side (the sender’s Oracle system)
We have not been able to “replicate” this problem using debug tools (identical data). Then it works fine and the connection is reset differently, i.e. WM waits for an ACK (after the END-OF-DATA package etc) before terminating the session.
My question is how this behavior can be controlled? We have discovered differences using SSL or not, using http1.0 or http1.1 etc. But we have not been able to pinpoint why WM cuts the connection so quickly when the Oracle system is sending data to us. I assume the answer is in the http protocol somewhere…
We use webMethods 6.5 and Trading Networks. An “HttpHeader” object is used in a java service to create the http response message. Basically setting the response code and adding http parameters.