[
https://issues.apache.org/jira/browse/FLINK-10026?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Till Rohrmann updated FLINK-10026:
----------------------------------
Fix Version/s: (was: 1.7.0)
> Integrate heap based timers in RocksDB backend with incremental checkpoints
> ---------------------------------------------------------------------------
>
> Key: FLINK-10026
> URL: https://issues.apache.org/jira/browse/FLINK-10026
> Project: Flink
> Issue Type: Sub-task
> Components: State Backends, Checkpointing
> Affects Versions: 1.6.0
> Reporter: Stefan Richter
> Priority: Major
> Fix For: 1.8.0
>
>
> Currently, heap based priority queue state does not participate in RocksDB
> incremental checkpointing and still uses the legacy (synchronous) way of
> checkpointing timers.
> We should integrate it with incremental checkpoints in such way that the
> state is written non-incrementally into a file, and another key-grouped state
> handle becomes part of the non-shared section of the incremental state handle.
> After that we can also remove the path for legacy checkpointing completely
> and also remove the key-grouping from the de-duplication hash map in the
> {{HeapPriorityQueueSet}}. All can go into one map.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
