Misses out on node events while a CI run is happening
The broker should ideally read node events in its own thread and insert them into a database, and also allow more than one (up to a configurable limit) CI run to happen at once.
The broker should ideally read node events in its own thread and insert them into a database, and also allow more than one (up to a configurable limit) CI run to happen at once.
This is an issue we are currently facing when an adapter requires too much time to process an event. During that period, the broker wont process any new event which might end up with hight latency on serving them.
The events are not lost unless the broker exits for some reason. So, it would also make sense to store them and process even the broker exits.