[
https://issues.apache.org/jira/browse/FLINK-4923?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15607949#comment-15607949
]
ASF GitHub Bot commented on FLINK-4923:
---------------------------------------
Github user StephanEwen commented on a diff in the pull request:
https://github.com/apache/flink/pull/2693#discussion_r85079795
--- Diff:
flink-runtime/src/main/java/org/apache/flink/runtime/io/network/partition/PipelinedSubpartition.java
---
@@ -220,4 +220,10 @@ boolean registerListener(NotificationListener
listener) {
throw new IllegalStateException("Already registered
listener.");
}
}
+
+ public int getNumberOfQueuedBuffers() {
+ synchronized (buffers) {
--- End diff --
Can we avoid synchronization here? The metrics should never influence
(block) the other code.
I would rather have the metrics be one off once in a while and avoid the
lock here.
> Expose input/output buffers and bufferPool usage as a metric for a Task
> -----------------------------------------------------------------------
>
> Key: FLINK-4923
> URL: https://issues.apache.org/jira/browse/FLINK-4923
> Project: Flink
> Issue Type: Improvement
> Components: Metrics
> Reporter: zhuhaifeng
> Assignee: zhuhaifeng
> Priority: Minor
> Fix For: 1.2.0
>
>
> We should expose the following Metrics on the TaskIOMetricGroup:
> 1. Buffers.numIn: received buffers of InputGates for a task
> 2. Buffers.numOut: buffers of produced ResultPartitions for a task
> 3. Buffers.InPoolUsage: usage of InputGates buffer pool for a task
> 4. Buffers.OutPoolUsage: usage of produced ResultPartitions buffer pool for
> a task
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
