Read a SIP trace: where to look first

A trace from a busy gateway holds hundreds of messages. Four questions get you to the cause faster than reading from the top.

1. Which messages belong to the call?

Find the INVITE for the number or time in question and note its Call-ID. Every message in that dialog carries it. A gateway or SBC that sits in the middle creates a second call leg with a different Call-ID, so find both legs before drawing conclusions.

2. Who ended it, and with what?

Look for the final response to the INVITE, or the BYE or CANCEL.

3. What did each side offer for media?

Read the SDP in the INVITE and in the answer (the 200 OK, or a 183).

4. What is different from a call that works?

Take a working call through the same path and compare the same message in each. Ignore fields that always differ, such as tags and branch values, and look at the request line, the identity headers and the SDP.

Collecting a clean trace on a CUBE

Set service timestamps debug datetime msec and a large logging buffered size, run debug ccsip messages, reproduce one call, then turn debugging off. One call in the buffer is easier to read than fifty.

A signaling trace shows what was agreed. It does not prove audio flowed. For that you need a packet capture.

Where Workbench helps

SIP Trace Analyzer draws the ladder and marks the failure. Call Comparison puts the working and failing calls side by side and highlights the fields that differ.