[ 
https://issues.apache.org/jira/browse/FLINK-26746?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18126263#comment-18126263
 ] 

ChenZhuo commented on FLINK-26746:
----------------------------------

Hi, I would like to work on this issue. Could a committer please assign it to 
me (Jira username: chenzhuo)?

Proposed scope: update the English and Chinese ProcessFunction timer 
fault-tolerance notes, remove the promise of a future fix via FLINK-10026 
(closed as Won’t Do), and correct the Chinese timer terminology.

In the current master implementation, 
RocksDBKeyedStateBackend.requiresLegacySynchronousTimerSnapshots returns true 
for heap timers and non-savepoint snapshots, regardless of whether incremental 
checkpointing is enabled. StreamOperatorStateHandler then writes these timers 
synchronously to raw keyed state. I therefore propose documenting the exception 
as RocksDB with heap-based timers, without restricting it to incremental 
checkpoints. Timers stored in RocksDB can use asynchronous checkpointing.

I will keep this documentation-only and validate the English and Chinese Hugo 
previews. Does this scope sound appropriate? I will wait for assignment and 
agreement before implementing.

> Update the documentation of Timer's Fault Tolerance section
> -----------------------------------------------------------
>
>                 Key: FLINK-26746
>                 URL: https://issues.apache.org/jira/browse/FLINK-26746
>             Project: Flink
>          Issue Type: Improvement
>          Components: Documentation
>            Reporter: Liwei Lin
>            Priority: Minor
>
> Currently in the documentation of Timer's Fault Tolerance section ( 
> [https://nightlies.apache.org/flink/flink-docs-release-1.14/docs/dev/datastream/operators/process_function/#fault-tolerance]
>  ), it says:
> Timers are always asynchronously checkpointed, except for the combination of 
> RocksDB backend / with incremental snapshots / with heap-based timers (will 
> be resolved with {{{}FLINK-10026{}}}). 
> Actually FLINK-10026 is resolved as 'Won't Do'; the documentation might be 
> updated.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to