pnowojski commented on a change in pull request #17229:
URL: https://github.com/apache/flink/pull/17229#discussion_r739165161
##########
File path:
flink-streaming-java/src/main/java/org/apache/flink/streaming/runtime/tasks/StreamTask.java
##########
@@ -523,11 +534,15 @@ protected void
processInput(MailboxDefaultAction.Controller controller) throws E
if (!recordWriter.isAvailable()) {
timer = new
GaugePeriodTimer(ioMetrics.getBackPressuredTimePerSecond());
resumeFuture = recordWriter.getAvailableFuture();
- } else {
+ } else if (!inputProcessor.isAvailable()) {
timer =
new ThroughputPeriodTimer(
ioMetrics.getIdleTimeMsPerSecond(),
throughputCalculator);
resumeFuture = inputProcessor.getAvailableFuture();
+ } else {
+ // todo: add new metrics (FLINK-23486)
+ timer = new
GaugePeriodTimer(ioMetrics.getBackPressuredTimePerSecond());
+ resumeFuture =
changelogWriterAvailabilityProvider.getAvailableFuture();
Review comment:
I think what you are proposing makes sense and the option (1) would be
cleaner and in the spirit of unix style cpu usage reporting (where everything
sums up to 100%). But this would most likely require _some_ Web UI changes:
1. Ideally we would need to find a good way to present the 4th task usage
time metric. It might be easy, but maybe we will struggle to find a space to do
it easily.
2. To limit Web UI changes, at the very least we could have a separate
metric for network and state backend based backpressure, but somewhere we could
add up those two values. For example either on the REST API level, or in the
Web UI itself. (probably in WebUI would be a better option).
--
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]
