[
https://issues.apache.org/jira/browse/SPARK-48628?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ziqi Liu updated SPARK-48628:
-
Description:
Currently there is no task on/off heap execution memory metrics. There is a
[peakExecutionMemory|https://github.com/apache/spark/blob/3cd35f8cb6462051c621cf49de54b9c5692aae1d/core/src/main/scala/org/apache/spark/executor/TaskMetrics.scala#L114]
metrics, however, the semantic is a confusing: it only cover the execution
memory used by shuffle/join/aggregate/sort, which is accumulated in specific
operators.
We can easily maintain the whole task-level peak memory in TaskMemoryManager,
assuming *acquireExecutionMemory* is the only one narrow waist for acquiring
execution memory.
Also it's nice to cleanup/deprecate that poorly-named `peakExecutionMemory`.
Creating two followup sub tickets:
* https://issues.apache.org/jira/browse/SPARK-48788 :accumulate task metrics
in stage, and display in Spark UI
* https://issues.apache.org/jira/browse/SPARK-48789 : deprecate
`peakExecutionMemory` once we have replacement for it.
was:
Currently there is no task on/off heap execution memory metrics. There is a
[peakExecutionMemory|https://github.com/apache/spark/blob/3cd35f8cb6462051c621cf49de54b9c5692aae1d/core/src/main/scala/org/apache/spark/executor/TaskMetrics.scala#L114]
metrics, however, the semantic is a bit confusing: it only cover the
execution memory used by shuffle/join/aggregate/sort, which is accumulated in
specific operators.
We can easily maintain the whole task-level peak memory in TaskMemoryManager,
assuming *acquireExecutionMemory* is the only one narrow waist for acquiring
execution memory.
> Add task peak on/off heap execution memory metrics
> --
>
> Key: SPARK-48628
> URL: https://issues.apache.org/jira/browse/SPARK-48628
> Project: Spark
> Issue Type: Improvement
> Components: Spark Core
>Affects Versions: 4.0.0
>Reporter: Ziqi Liu
>Priority: Major
>
> Currently there is no task on/off heap execution memory metrics. There is a
> [peakExecutionMemory|https://github.com/apache/spark/blob/3cd35f8cb6462051c621cf49de54b9c5692aae1d/core/src/main/scala/org/apache/spark/executor/TaskMetrics.scala#L114]
> metrics, however, the semantic is a confusing: it only cover the execution
> memory used by shuffle/join/aggregate/sort, which is accumulated in specific
> operators.
>
> We can easily maintain the whole task-level peak memory in TaskMemoryManager,
> assuming *acquireExecutionMemory* is the only one narrow waist for acquiring
> execution memory.
>
> Also it's nice to cleanup/deprecate that poorly-named `peakExecutionMemory`.
>
> Creating two followup sub tickets:
> * https://issues.apache.org/jira/browse/SPARK-48788 :accumulate task metrics
> in stage, and display in Spark UI
> * https://issues.apache.org/jira/browse/SPARK-48789 : deprecate
> `peakExecutionMemory` once we have replacement for it.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org