flink-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stefan Richter <s.rich...@data-artisans.com>
Subject Re: Managed operator state treating state of all parallel operators as the same
Date Tue, 04 Jul 2017 13:19:47 GMT
What I mean is that you could obtain such a state in 

initializeState(FunctionInitializationContext context) {

and in snapshotState(…), you will just insert the state in only one of the parallel instances.
Which instance can be based on the subtask index (e.g. only add to the list state if your
subtask index is == 0). You can obtain the subtask index by getRuntimeContext().getIndexOfThisSubtask().

UnionListState will replicate all submitted states to all parallel instances, so what was
checkpointed on one operator instance will be replicated to all in restore.


> Am 04.07.2017 um 13:56 schrieb gerardg <gerard@talaia.io>:
> Thanks Fabian, I'll keep an eye to that JIRA.
> I'm not sure I follow you Stefan. You mean that I could implement my own
> OperatorStateStore and override its methods (e.g. snapshot and restore) to
> achieve this functionality? I think I don't have enough knowledge about
> Flink's internals to implement this easily.
> Gerard
> --
> View this message in context: http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/Managed-operator-state-treating-state-of-all-parallel-operators-as-the-same-tp14102p14111.html
> Sent from the Apache Flink User Mailing List archive. mailing list archive at Nabble.com.

View raw message