wypoon commented on issue #25659: [SPARK-28770][CORE][TESTS]Ignore 
SparkListenerStageExecutorMetrics in testApplicationReplay test
URL: https://github.com/apache/spark/pull/25659#issuecomment-527741025
 
 
   `SparkListenerStageExecutorMetrics` were introduced in SPARK-23429 
(https://github.com/apache/spark/pull/21221/). By design, executor metrics 
update events are not logged by the `EventLoggingListener`. Instead, the 
listener keeps track of the per stage peaks for any of the executors and the 
driver for which it has received metrics. On stage completion, the peaks for 
the stage are logged in `SparkListenerStageExecutorMetrics` events for each of 
these executors and driver.
   Since executor metrics update events are not logged in the event log, they 
do not get replayed. Thus the listener for the replay never sees metrics 
updates. It is therefore valid to exclude `SparkListenerStageExecutorMetrics` 
events from both the original and the replay for the purpose of comparison.
   
   However, instead of excluding all `SparkListenerStageExecutorMetrics` events 
from both the original `EventLoggingListener` and the replay listener, we can 
have a finer-grained fix, which I have proposed in 
https://github.com/apache/spark/pull/25673/ for comparison. It should be 
sufficient to exclude any `SparkListenerStageExecutorMetrics` events for the 
driver. This is because with SPARK-26329 
(https://github.com/apache/spark/pull/23767/), executor metrics are also sent 
in task end events (which do get replayed), so the `EventLoggingListener` 
always receives metrics for the executors (just not necessarily for the 
driver), and thus `SparkListenerStageExecutorMetrics` events for the executors 
always get logged.
   
   

----------------------------------------------------------------
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.
 
For queries about this service, please contact Infrastructure at:
[email protected]


With regards,
Apache Git Services

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to