We write a service that runs on the 7970 phones, and have noticed that on occasion the phone hangs on a request to our server.
It will clear itself in about 20 or so seconds, but we are trying to improve the performance.
This is reproducible in the field with call managers with large systems, and on our local test system where we use the call manager simulator directly attached to the same switch as the application server it is communicating with.
We have had no problem with these same requests being made from browsers over the same network - there are multiple interfaces that reference the same application, full browser based, pda "simple web" based, and phone based - the phone is the only problem point. We are confident that our server is responding in a timely fashion.
Wireshark was used to capture packets and the results of three hangs have been condensed into one file with a couple of extra lines of info on each side of the incident to make sure that the entire sequence is captured - in truth this was from a much longer test period, but if you look at times the grouping is obvious - 1-11, 12-23 and 24 to 34. That file is attached. 220.127.116.11 is the application server running tomcat, and 18.104.22.168 is the IP Phone in the attached file.
What happens is that [SYN] commands are repeatedly sent, about 1.5 seconds apart, they fail about a half dozen times, then either the user represses the button - which is seen in the first two hangs in the file. If the request times out, the sequence number increases, and the request is passed.
We would be very interested in hearing suggestions on how to fine tune the system, and any information on why requests seem to go into this blocked state. To our untrained eyes it looks like the phone is not receiving an [ACK] reply that it expects, but we could use some low level TCP-IP advice on this.