Error monitoring

[tim-schilling]

Monitor errors via got_request_exception signal.

The middleware process_exception would only allow us to track
exceptions thrown in views, that weren't picked up by an
earlier middleware that returned a result. This solution allows
us to track exceptions that occur anywhere in the application.

The downside to this is that it's possible if another receiver
of got_request_exception throws a new exception, our handler
may get the wrong exception with sys.exc_info().

We could limit the scope of that issue by injecting a middleware
that's as far down the list as possible to wrap the view
and catch exceptions there. Then rely on the signal handler
for catching exceptions in the middleware.

Create ErrorServiceThread to buffer error messages.

TODO:

• Django instrumentation unit tests
• ErrorMonitor core unit tests
• Documentation of new settings and error monitoring feature.
• Batch errors together and send on separate thread from the requests thread.
• API for manually tracking errors

[dlanderson]

Where does ErrorServiceThread get started? Does it start even if errors_enabled config is false?

[tim-schilling]

Where does ErrorServiceThread get started? Does it start even if errors_enabled config is false?

ErrorServiceThread.ensure_started is called at the end of ErrorServiceThread.send. This keeps the thread from being created until it's needed.

[dlanderson]

@tim-schilling If we've tested and all looks good and you're ready to release let's go ahead 👍

[tim-schilling]

I'm going to push the existing changes out in a version, then merge this as another version.