trident parallelization and failover

[Weide Zhang <we...@gmail.com>]

Hi,

If i have specify parallelHint for TridentTransactionalKafkaSpout as more
than 1, internally, what's the class of the instance of trident bolt is
created more than 1 times ? Is it tridentspoutexecutor or
tridentspoutcoordinator or both ?  Can they run on different worker node ?

Also, it seems either masterbatchcoordinator and tridentspoutcoordinator
manages the whole topology status.  what if they fail and nimbus restarted
worker node? will we lose information and will exact once semantic still be
achieved ?

Thanks,

Weide