[jira] [Commented] (SPARK-21501) Spark shuffle index cache size should be memory based
[ https://issues.apache.org/jira/browse/SPARK-21501?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17218285#comment-17218285 ] Lars Francke commented on SPARK-21501: -- Just FYI for others stumbling across this: This has a bug in how the memory is calculated and might use way more than the 100MB it intends to. See SPARK-33206 for details. > Spark shuffle index cache size should be memory based > - > > Key: SPARK-21501 > URL: https://issues.apache.org/jira/browse/SPARK-21501 > Project: Spark > Issue Type: Bug > Components: Shuffle, Spark Core >Affects Versions: 2.1.0 >Reporter: Thomas Graves >Assignee: Sanket Reddy >Priority: Major > Fix For: 2.3.0 > > > Right now the spark shuffle service has a cache for index files. It is based > on a # of files cached (spark.shuffle.service.index.cache.entries). This can > cause issues if people have a lot of reducers because the size of each entry > can fluctuate based on the # of reducers. > We saw an issues with a job that had 17 reducers and it caused NM with > spark shuffle service to use 700-800MB or memory in NM by itself. > We should change this cache to be memory based and only allow a certain > memory size used. When I say memory based I mean the cache should have a > limit of say 100MB. -- 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
[jira] [Commented] (SPARK-21501) Spark shuffle index cache size should be memory based
[ https://issues.apache.org/jira/browse/SPARK-21501?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16362424#comment-16362424 ] Xun REN commented on SPARK-21501: - Hi guys, Could you tell me how to figure out how many memory the NM with spark shuffle service has used ? And how to know a spark job has used how many reducers ? Because I have the same problem recently and I want to get a list of spark jobs by sorting by number of reducers. Thanks. Regards, Xun REN. > Spark shuffle index cache size should be memory based > - > > Key: SPARK-21501 > URL: https://issues.apache.org/jira/browse/SPARK-21501 > Project: Spark > Issue Type: Bug > Components: Shuffle >Affects Versions: 2.1.0 >Reporter: Thomas Graves >Assignee: Sanket Reddy >Priority: Major > Fix For: 2.3.0 > > > Right now the spark shuffle service has a cache for index files. It is based > on a # of files cached (spark.shuffle.service.index.cache.entries). This can > cause issues if people have a lot of reducers because the size of each entry > can fluctuate based on the # of reducers. > We saw an issues with a job that had 17 reducers and it caused NM with > spark shuffle service to use 700-800MB or memory in NM by itself. > We should change this cache to be memory based and only allow a certain > memory size used. When I say memory based I mean the cache should have a > limit of say 100MB. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-21501) Spark shuffle index cache size should be memory based
[ https://issues.apache.org/jira/browse/SPARK-21501?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16100658#comment-16100658 ] Sanket Reddy commented on SPARK-21501: -- Hi I am working on this issue just to avoid any redundancies if any thanks > Spark shuffle index cache size should be memory based > - > > Key: SPARK-21501 > URL: https://issues.apache.org/jira/browse/SPARK-21501 > Project: Spark > Issue Type: Bug > Components: Shuffle >Affects Versions: 2.1.0 >Reporter: Thomas Graves > > Right now the spark shuffle service has a cache for index files. It is based > on a # of files cached (spark.shuffle.service.index.cache.entries). This can > cause issues if people have a lot of reducers because the size of each entry > can fluctuate based on the # of reducers. > We saw an issues with a job that had 17 reducers and it caused NM with > spark shuffle service to use 700-800MB or memory in NM by itself. > We should change this cache to be memory based and only allow a certain > memory size used. When I say memory based I mean the cache should have a > limit of say 100MB. -- This message was sent by Atlassian JIRA (v6.4.14#64029) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-21501) Spark shuffle index cache size should be memory based
[ https://issues.apache.org/jira/browse/SPARK-21501?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=1616#comment-1616 ] Thomas Graves commented on SPARK-21501: --- We want to change it from a # of entries to a size of memory. So the cache is enabled but has a max size of 200MB for instance. This way you can size that based on the memory size of the YARN NM where the spark shuffle service runs. If I have a 1GB heap on my NM this ensures it only uses 200MB. With the # of entries you can't guarantee it won't use all 1GB of your heap because the size of each entry is dependent upon the # of reducers in that job. > Spark shuffle index cache size should be memory based > - > > Key: SPARK-21501 > URL: https://issues.apache.org/jira/browse/SPARK-21501 > Project: Spark > Issue Type: Bug > Components: Shuffle >Affects Versions: 2.1.0 >Reporter: Thomas Graves > > Right now the spark shuffle service has a cache for index files. It is based > on a # of files cached (spark.shuffle.service.index.cache.entries). This can > cause issues if people have a lot of reducers because the size of each entry > can fluctuate based on the # of reducers. > We saw an issues with a job that had 17 reducers and it caused NM with > spark shuffle service to use 700-800MB or memory in NM by itself. > We should change this cache to be memory based and only allow a certain > memory size used. When I say memory based I mean the cache should have a > limit of say 100MB. -- This message was sent by Atlassian JIRA (v6.4.14#64029) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-21501) Spark shuffle index cache size should be memory based
[
https://issues.apache.org/jira/browse/SPARK-21501?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16099351#comment-16099351
]
Kazuaki Ishizaki commented on SPARK-21501:
--
I see. I misunderstood the description.
You expect that memory cache would be enabled even when # of entries is larger
than {{spark.shuffle.service.index.cache.entries}} if the total cache size is
not large.
> Spark shuffle index cache size should be memory based
> -
>
> Key: SPARK-21501
> URL: https://issues.apache.org/jira/browse/SPARK-21501
> Project: Spark
> Issue Type: Bug
> Components: Shuffle
>Affects Versions: 2.1.0
>Reporter: Thomas Graves
>
> Right now the spark shuffle service has a cache for index files. It is based
> on a # of files cached (spark.shuffle.service.index.cache.entries). This can
> cause issues if people have a lot of reducers because the size of each entry
> can fluctuate based on the # of reducers.
> We saw an issues with a job that had 17 reducers and it caused NM with
> spark shuffle service to use 700-800MB or memory in NM by itself.
> We should change this cache to be memory based and only allow a certain
> memory size used. When I say memory based I mean the cache should have a
> limit of say 100MB.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-21501) Spark shuffle index cache size should be memory based
[ https://issues.apache.org/jira/browse/SPARK-21501?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16098907#comment-16098907 ] Thomas Graves commented on SPARK-21501: --- The issue was actually introduced with SPARK-15074. Updated the affects version. I was thinking that was added earlier so thanks for pointing out. That cache should be memory based not # of entries based. > Spark shuffle index cache size should be memory based > - > > Key: SPARK-21501 > URL: https://issues.apache.org/jira/browse/SPARK-21501 > Project: Spark > Issue Type: Bug > Components: Shuffle >Affects Versions: 2.1.0 >Reporter: Thomas Graves > > Right now the spark shuffle service has a cache for index files. It is based > on a # of files cached (spark.shuffle.service.index.cache.entries). This can > cause issues if people have a lot of reducers because the size of each entry > can fluctuate based on the # of reducers. > We saw an issues with a job that had 17 reducers and it caused NM with > spark shuffle service to use 700-800MB or memory in NM by itself. > We should change this cache to be memory based and only allow a certain > memory size used. When I say memory based I mean the cache should have a > limit of say 100MB. -- This message was sent by Atlassian JIRA (v6.4.14#64029) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-21501) Spark shuffle index cache size should be memory based
[ https://issues.apache.org/jira/browse/SPARK-21501?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16098531#comment-16098531 ] Kazuaki Ishizaki commented on SPARK-21501: -- I guess that to use Spark 2.1 or later version alleviates this issue by SPARK-15074 > Spark shuffle index cache size should be memory based > - > > Key: SPARK-21501 > URL: https://issues.apache.org/jira/browse/SPARK-21501 > Project: Spark > Issue Type: Bug > Components: Shuffle >Affects Versions: 2.0.0 >Reporter: Thomas Graves > > Right now the spark shuffle service has a cache for index files. It is based > on a # of files cached (spark.shuffle.service.index.cache.entries). This can > cause issues if people have a lot of reducers because the size of each entry > can fluctuate based on the # of reducers. > We saw an issues with a job that had 17 reducers and it caused NM with > spark shuffle service to use 700-800MB or memory in NM by itself. > We should change this cache to be memory based and only allow a certain > memory size used. When I say memory based I mean the cache should have a > limit of say 100MB. -- This message was sent by Atlassian JIRA (v6.4.14#64029) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
