[
https://issues.apache.org/jira/browse/FLINK-40524?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Rui Fan resolved FLINK-40524.
-----------------------------
Fix Version/s: 2.4.0
Resolution: Fixed
Merged to master(2.4.0) via: 9bd05c557c74c21e68faa6f5a11e47c6c513c025
> LocalInputChannel.getNextBuffer takes the recoveredBuffers monitor on every
> buffer even when the channel never needed recovery
> ------------------------------------------------------------------------------------------------------------------------------
>
> Key: FLINK-40524
> URL: https://issues.apache.org/jira/browse/FLINK-40524
> Project: Flink
> Issue Type: Sub-task
> Components: Runtime / Checkpointing
> Reporter: Rui Fan
> Assignee: Rui Fan
> Priority: Major
> Labels: pull-request-available
> Fix For: 2.4.0
>
>
> CDR added an unconditional `synchronized (recoveredBuffers)` to the
> local-consume hot path [1], but for channels built with `needsRecovery=false`
> (all local channels when CDR is off) `inRecovery` is always false and
> `recoveredBuffers` always empty, so the block is dead yet still costs a
> monitor acquire per buffer. Fix: since `needsRecovery` is final, gate the
> recovery block on it so the steady-state path skips the monitor entirely.
> [1]
> https://github.com/apache/flink/blob/76e774e04ff84ea305714153c5f4790fff98437c/flink-runtime/src/main/java/org/apache/flink/runtime/io/network/partition/consumer/LocalInputChannel.java#L566
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
