Actual behavior: The worker dies, because its error handler itself throws an error when trying to reset the job because it can't communicate with mongo
When this happens, that Qu instance's aggregation worker is effectively dead and will not process additional aggregations. This results in a buildup of aggregations until that Qu instance is restarted
Expected behavior:
The error handler should not throw an error, or, if it does, the worker should be restarted
The worker's currently processing job, at the time of failure, would get restarted when a) the worker is restarted and b) able to communicate with mongo successfully
Actual behavior: The worker dies, because its error handler itself throws an error when trying to reset the job because it can't communicate with mongo
When this happens, that Qu instance's aggregation worker is effectively dead and will not process additional aggregations. This results in a buildup of aggregations until that Qu instance is restarted
Expected behavior: