You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Stephan Ewen (JIRA)" <ji...@apache.org> on 2019/08/01 08:22:00 UTC
[jira] [Updated] (FLINK-13326) Support asynchronous writing to raw
operator (and raw keyed) state
[ https://issues.apache.org/jira/browse/FLINK-13326?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Stephan Ewen updated FLINK-13326:
---------------------------------
Fix Version/s: 1.9.0
> Support asynchronous writing to raw operator (and raw keyed) state
> ------------------------------------------------------------------
>
> Key: FLINK-13326
> URL: https://issues.apache.org/jira/browse/FLINK-13326
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / State Backends
> Reporter: Igal Shilman
> Assignee: Igal Shilman
> Priority: Minor
> Labels: pull-request-available
> Fix For: 1.9.0, 1.10.0
>
> Time Spent: 20m
> Remaining Estimate: 0h
>
> Currently in AbstractStreamOperator, it is expected that the implementors of snapshotState, that use either raw operator state or raw keyed operator state, would complete the snapshotting of their state in the synchronous part of the checkpoint (before the snapshotState method complete)
> But sometimes it is desirable to snapshot custom data structure asynchronously (i.e. out side of the snapshotState method) in a similar way that the other state backends allow.
> This issue is about adding the ability to complete the raw states outside of the snapshotState method, and hence effectively support async snapshots for raw states.
> Flink snapshotting logic already supports splitting the snapshot to sync and async parts for the various state backends, therefore the expected changes mainly to reuse that logic for the raw keyed/ raw operator states.
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)