hive-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sergey Shelukhin (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HIVE-10103) LLAP: Cancelling tasks fails to stop cache filling threads
Date Wed, 01 Apr 2015 01:23:52 GMT

    [ https://issues.apache.org/jira/browse/HIVE-10103?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14389832#comment-14389832
] 

Sergey Shelukhin commented on HIVE-10103:
-----------------------------------------

Main problem is that tez doesn't close the reader when operators fail... trying to see now
how to change that. Added some workarounds for when nextCvb is interrupted for now


> LLAP: Cancelling tasks fails to stop cache filling threads
> ----------------------------------------------------------
>
>                 Key: HIVE-10103
>                 URL: https://issues.apache.org/jira/browse/HIVE-10103
>             Project: Hive
>          Issue Type: Sub-task
>            Reporter: Gopal V
>            Assignee: Sergey Shelukhin
>
> Running a bad query (~1Tb scan on a 1Gb cache) and killing the tasks via the container
launcher fails to free up the cache filler threads.
> The cache filler threads with no consumers get stuck into a loop 
> {code}
> 2015-03-26 14:02:47,335 [pool-2-thread-2(container_222212222_1659_01_000074_gopal_20150326135614_2bb61f02-3c2b-4512-a34e-81803cd13fb6:1_Map
1_73_0)] WARN org.apache.hadoop.hive.llap.io.api.impl.LlapIoImpl: Cannot evict blocks for
262144 calls; cache full?
> 2015-03-26 14:02:48,018 [pool-2-thread-7(container_222212222_1659_01_000076_gopal_20150326135614_2bb61f02-3c2b-4512-a34e-81803cd13fb6:1_Map
1_75_0)] WARN org.apache.hadoop.hive.llap.io.api.impl.LlapIoImpl: Cannot evict blocks for
262144 calls; cache full?
> 2015-03-26 14:02:51,658 [pool-2-thread-1(container_222212222_1659_01_000073_gopal_20150326135614_2bb61f02-3c2b-4512-a34e-81803cd13fb6:1_Map
1_72_0)] WARN org.apache.hadoop.hive.llap.io.api.impl.LlapIoImpl: Cannot evict blocks for
262144 calls; cache full?
> {code}
> Needs to kill a daemon to get back to normal operations.



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

Mime
View raw message