In an AWS cluster, one of the great things is that you can scale out
automatically when the load increases. The CLI for storm allows for
rebalancing after I have brought up my newly provisioned nodes, but my
question relates to system availability during this time.
Am I correct in saying that workers stop during this time? I.e. if I wanted
to make the system seem available I would have to ensure this time is
sufficiently small and use Asynch comms, but I have no way of making it
100% available during this time?
What has your approach been to upgrades in respect of availability?