[
https://issues.apache.org/jira/browse/HIVE-18652?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16444287#comment-16444287
]
Sahil Takiar commented on HIVE-18652:
-------------------------------------
Now the console logs show this:
{code}
Spark Job[1] Metrics: TaskDurationTime: 570 ExecutorCpuTime: 390
JvmGCTime: 0 BytesRead / RecordsRead: 11150 / 500 ShuffleTotalBytesRead /
ShuffleRecordsRead: 2445 / 1 ShuffleBytesWritten / ShuffleRecordsWritten:
2445 / 1
{code}
I decided to expose only these metrics because they are the ones that are
exposed by default on the Spark Web UI.
Some follow up enhancements:
* HIVE-19051: Add units to the displayed metrics
* HIVE-19176: When we implement this, we can also add metrics for each
individual Spark stage, right now the granularity is at the job level
> Print Spark metrics on console
> ------------------------------
>
> Key: HIVE-18652
> URL: https://issues.apache.org/jira/browse/HIVE-18652
> Project: Hive
> Issue Type: Sub-task
> Components: Spark
> Reporter: Sahil Takiar
> Assignee: Sahil Takiar
> Priority: Major
> Attachments: HIVE-18652.1.patch, HIVE-18652.2.patch
>
>
> For Hive-on-MR, each MR job launched prints out some stats about the job:
> {code}
> INFO : 2018-02-07 17:51:11,218 Stage-1 map = 0%, reduce = 0%
> INFO : 2018-02-07 17:51:18,396 Stage-1 map = 100%, reduce = 0%, Cumulative
> CPU 1.87 sec
> INFO : 2018-02-07 17:51:25,742 Stage-1 map = 100%, reduce = 100%,
> Cumulative CPU 4.34 sec
> INFO : MapReduce Total cumulative CPU time: 4 seconds 340 msec
> INFO : Ended Job = job_1517865654989_0004
> INFO : MapReduce Jobs Launched:
> INFO : Stage-Stage-1: Map: 1 Reduce: 1 Cumulative CPU: 4.34 sec HDFS
> Read: 7353 HDFS Write: 151 SUCCESS
> INFO : Total MapReduce CPU Time Spent: 4 seconds 340 msec
> {code}
> We should do the same for Hive-on-Spark.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
