[
https://issues.apache.org/jira/browse/FLINK-10146?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17237390#comment-17237390
]
Nico Kruber commented on FLINK-10146:
-------------------------------------
It looks like it. It reflects the number of buffers that (eventually) go out of
the {{RecordWriter}} but not the number of buffers sent over network.
> numBuffersOut not updated properly
> ----------------------------------
>
> Key: FLINK-10146
> URL: https://issues.apache.org/jira/browse/FLINK-10146
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Metrics, Runtime / Network
> Affects Versions: 1.5.3, 1.6.1, 1.7.0
> Reporter: Nico Kruber
> Assignee: Nico Kruber
> Priority: Major
>
> We only enqueue a {{BufferConsumer}} once into the result (sub)partition but
> Netty may draw and send parts of it independently. {{numBuffersOut}} is,
> however, only increased when a new buffer is retrieved from the buffer pool,
> not when an actual buffer is sent via Netty. This metric is thus wrong.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
