Logging prefix for call chains on a distributed system

[EliAndrewC]

Moved here from the original (now private) issue tracker.

As a developer trying to debug a problem in production, I'd like a common logging prefix to be shown on all related requests across multiple servers, to help me isolate relevant log messages.

Let's say a user on example.com makes a websocket request to foo.bar. Behind the scenes, that method makes a separate request to baz.baf, which is a remote service running on example.org, so we make an websocket RPC request.

When debugging a problem, it can be tricky and annoying to link the requests on different servers to one another. Since all logging goes through sideboard.lib.log, it would be pretty easy to generate a random logging token and prepend that to our log messages. When making any outgoing request, we'd add a log_token field to the websocket JSON, which would be used if present.

[binary1230]

YEA, also awesome. I was thinking about the exact same thing for cherrypy, implementing some kind of request_id for logging

[EliAndrewC]

Agreed, I didn't necessarily make this clear in the ticket description, but I intend this as a general feature and not just for websocket call chains.