infoverload commented on a change in pull request #472:
URL: https://github.com/apache/flink-web/pull/472#discussion_r719428381
##########
File path: _posts/2021-09-29-release-1.14.0.md
##########
@@ -163,12 +163,11 @@ Apache Flink buffers a certain amount of data in its
network stack to be able to
bandwidth of fast networks. A Flink application running with high throughput
uses some (or
all) of that memory. Aligned checkpoints flow with the data through the
network buffers in milliseconds.
-When a Flink application becomes (temporarily) backpressured (for example when
being backpressured
-by an external system, or when hitting skewed records), this typically results
in a lot more data in
-the network buffers than is necessary utilize enough network bandwith for the
application's current throughput
-(which is lowered due to backpressure). There is even an adverse effect: more
buffered data means
-the checkpoints need to do more work. Aligned checkpoint barriers need to wait
for more data to be
-processed, unaligned checkpoints need to persist more in-flight data.
+During (temporal) backpressure from a resource bottleneck such as an external
system, data skew, or (temporarily)
+increased load, Flink was buffering a lot more data inside its network buffers
than necessary to utilize
+enough network bandwith for the application's current – backpressured –
throughput. This actually has an adverse
+effect: more buffered data means the checkpoints need to do more work: Aligned
checkpoint barriers need to wait
+for more data to be processed, unaligned checkpoints need to persist more
in-flight data.
Review comment:
```suggestion
During (temporary) backpressure from a resource bottleneck such as an
external system, data skew, or (temporarily)
increased load, Flink was buffering a lot more data inside its network
buffers than necessary to utilize
enough network bandwidth for the application's current – backpressured –
throughput. This actually has an adverse
effect because more buffered data means that the checkpoints need to do more
work. Aligned checkpoint barriers need to wait for more data to be processed,
unaligned checkpoints need to persist more in-flight data.
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
