[
https://issues.apache.org/jira/browse/SPARK-19659?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15964476#comment-15964476
]
Imran Rashid commented on SPARK-19659:
--------------------------------------
currently the smallest unit is an entire block. That certainly could be
changed, but that is another larger change to the way spark uses netty.
bq. If the unit is block, I think it's really hard to avoid OOM entirely, as if
the estimated block size is wrong, fetching this block may cause OOM and we can
do nothing about it.
I don't totally agree with this. Yes, the estimated block size could be wrong.
The main problem now, though, is that the error is totally unbounded. I
really don't think it should be hard to change spark to provide reasonable
bounds on that error. Even if that first step doesn't entirely eliminate the
chance of OOM, it would certainly help. And that can continue to be improved
upon.
> Fetch big blocks to disk when shuffle-read
> ------------------------------------------
>
> Key: SPARK-19659
> URL: https://issues.apache.org/jira/browse/SPARK-19659
> Project: Spark
> Issue Type: Improvement
> Components: Shuffle
> Affects Versions: 2.1.0
> Reporter: jin xing
> Attachments: SPARK-19659-design-v1.pdf, SPARK-19659-design-v2.pdf
>
>
> Currently the whole block is fetched into memory(offheap by default) when
> shuffle-read. A block is defined by (shuffleId, mapId, reduceId). Thus it can
> be large when skew situations. If OOM happens during shuffle read, job will
> be killed and users will be notified to "Consider boosting
> spark.yarn.executor.memoryOverhead". Adjusting parameter and allocating more
> memory can resolve the OOM. However the approach is not perfectly suitable
> for production environment, especially for data warehouse.
> Using Spark SQL as data engine in warehouse, users hope to have a unified
> parameter(e.g. memory) but less resource wasted(resource is allocated but not
> used),
> It's not always easy to predict skew situations, when happen, it make sense
> to fetch remote blocks to disk for shuffle-read, rather than
> kill the job because of OOM. This approach is mentioned during the discussion
> in SPARK-3019, by [~sandyr] and [~mridulm80]
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
