rabbitmq1 is frequently hitting 80-90% CPU usage. It's current running celerybeat,process_objects,cycle_data,calculate_eta,populate_performance_series,fetch_bugs - plus trying to act as the rabbitmq host. Now that bug 1122139 has split the default queue up, we should move several of these elsewhere (particularly populate_performance_series). I think this may be an aggravating factor with bug 1125410.

Less of an issue now we've updated to python 2.7 and gotten the json perf win.

This has been fine recently - and we'll likely be splitting the nodes/tasks up more for Heroku anyway.