That's reasonable. It's like making any other external database call. It might be a good idea to add batch DRPC requests to the API so you could reduce roundtrips (would require a patch to Storm), though you can scale it by adding more DRPC servers and more parallelism.
On Thu, Jul 19, 2012 at 3:31 PM, Calvin French-Owen
<cal...@segment.io> wrote:
There are a few pieces of my application which I'd like to either make synchronous or return a value which I can then process from other bolts. Is there any downside to making DRPC calls from within bolts in the topology other than that the thread will block?
What's generally the best way to do this - is DRPC the recommended approach? Can I just increase the parallelism of bolts which might be pausing for DRPC calls?
Thanks for any help - I'm new to storm but it seems like a very cool system.