You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@spark.apache.org by "Sean Owen (JIRA)" <ji...@apache.org> on 2015/11/17 11:51:10 UTC

[jira] [Resolved] (SPARK-11776) Cache fraction hardly reach 100%

     [ https://issues.apache.org/jira/browse/SPARK-11776?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Sean Owen resolved SPARK-11776.
-------------------------------
    Resolution: Invalid

I don't think this describes a valid problem. For example, there is no indication that you even have enough data to fill the cache.

> Cache fraction hardly reach 100%
> --------------------------------
>
>                 Key: SPARK-11776
>                 URL: https://issues.apache.org/jira/browse/SPARK-11776
>             Project: Spark
>          Issue Type: Bug
>            Reporter: Wenmin Wu
>
> Even with plenty partitions as well as executor memory the cache fraction can hardly reach 100%. What even worse is that those uncached executor has too much records. Here is the summary of the aggregated metrics:
> 1 	hd268dg.prod.mediav.com:19344 	23 s 	1 	0 	1 	339.5 MB / 314 	5.7 MB / 1
> 10 	hd268dg.prod.mediav.com:2342 	23 s 	1 	0 	1 	340.0 MB / 315 	5.7 MB / 1
> 11 	hd258dg.prod.mediav.com:63130 	23 s 	1 	0 	1 	340.1 MB / 315 	5.7 MB / 1
> 12 	hd278dg.prod.mediav.com:57601 	15 s 	1 	0 	1 	340.6 MB / 315 	5.7 MB / 1
> 13 	hd268dg.prod.mediav.com:38740 	25 s 	1 	0 	1 	340.0 MB / 315 	5.7 MB / 1
> 14 	hd258dg.prod.mediav.com:34498 	22 s 	1 	0 	1 	339.8 MB / 315 	5.7 MB / 1
> 16 	hd268dg.prod.mediav.com:27714 	25 s 	1 	0 	1 	341.1 MB / 316 	5.7 MB / 1
> 17 	hd258dg.prod.mediav.com:36584 	22 s 	1 	0 	1 	340.7 MB / 315 	5.7 MB / 1
> 18 	hd278dg.prod.mediav.com:37627 	16 s 	1 	0 	1 	340.8 MB / 315 	5.7 MB / 1
> 19 	hd268dg.prod.mediav.com:5938 	25 s 	1 	0 	1 	339.8 MB / 315 	5.7 MB / 1
> 2 	hd258dg.prod.mediav.com:46592 	2.3 min 	1 	0 	1 	190.8 MB / 628778 	6.5 MB / 1
> 20 	hd258dg.prod.mediav.com:29043 	2.2 min 	1 	0 	1 	190.5 MB / 627972 	6.5 MB / 1 
> You can see the records in executor 2 and executor 20 is too much with respect to the storage. I think some bugs must exists during caching stage. 



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org