Hey everybody
We have an issue with bolts that take a long time to prepare. Since Storm doesn't wait for the entire topology to be ready before emitting tuples we get tuples in the topology before they can actually be processed. We have a low latency situation going and we need to be able to start emitting from spout only when we can process stuff quickly enough. Is there a built-in way to track bolt prepares? My other options would be to: a) write prepares to redis or some other single source or b) pass a test tuple through my topology and make sure it acks before sending real data. Did anyone take any of these approaches? any insights I can use? Thanks!
