I think that the futures may be lost if the machine is down for some reasons.
So simply using the futures may not be acceptable in production.
I hope this is helpful.
> 在 2016年9月16日，下午3:01，alghimo <albert.gimenez.mora...@gmail.com> 写道：
> Hi Timo,
> Thank you for the fast reply :) I will check and follow the discussion, it
> would be really cool to be able to have that feature.
> As for the updates, there I'm using Futures which update accumulators on
> success / failure, and everything seems to be working fine, I tested both in
> local mode and deploying in YARN with multiple task managers, although they
> were all running on the same machine. But this update job has no sinks,
> maybe that's why it's working.
> Thanks again!
> View this message in context:
> Sent from the Apache Flink Mailing List archive. mailing list archive at