[
https://issues.apache.org/jira/browse/SPARK-31380?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17085138#comment-17085138
]
Srinivas Rishindra Pothireddi commented on SPARK-31380:
-------------------------------------------------------
I tested this with spark master at the time of creating this ticket. I am not
seeing this issue again now and spark-3.0.0-preview2. I guess the issue might
have gone away recently. The issue definitely exists with spark-2.4.5.
The history server might be flaky. That could be a reason why we are seeing
this issue intermittently. For example when I tried to run my application to
test it I am able to see the metrics in safari but not in chrome.
!Screen Shot 2020-04-16 at 10.55.17 AM.png!
!Screen Shot 2020-04-16 at 10.57.14 AM.png!
> Peak Execution Memory Quantile is not displayed in Spark History Server UI
> --------------------------------------------------------------------------
>
> Key: SPARK-31380
> URL: https://issues.apache.org/jira/browse/SPARK-31380
> Project: Spark
> Issue Type: Bug
> Components: Spark Core, Web UI
> Affects Versions: 3.0.0
> Reporter: Srinivas Rishindra Pothireddi
> Priority: Major
> Attachments: image-2020-04-15-18-16-18-254.png
>
>
> Peak Execution Memory Quantile is displayed in the regular Spark UI
> correctly. If the same application is viewed in Spark History Server UI, Peak
> Execution Memory is always displayed as zero.
> Spark event log for the application seem to contain Peak Execution
> Memory(under the tag "internal.metrics.peakExecutionMemory") correctly.
> However this is not reflected in the History Server UI.
> *Steps to produce non-zero Peak Execution Memory*
> spark.range(0, 200000).map\{x => (x , x % 20)}.toDF("a",
> "b").createOrReplaceTempView("fred")
> spark.range(0, 200000).map\{x => (x , x + 1)}.toDF("a",
> "b").createOrReplaceTempView("phil")
> sql("select p.**,* f.* from phil p join fred f on f.b = p.b").count
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
