hive-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Kai Sasaki (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HIVE-11749) To sometimes deadlock when run a query
Date Mon, 02 Nov 2015 07:51:27 GMT

    [ https://issues.apache.org/jira/browse/HIVE-11749?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14984852#comment-14984852
] 

Kai Sasaki commented on HIVE-11749:
-----------------------------------

It seems that {{HashTable}} from {{GlobalWorkMapFactory}} is not thread-local. {{GlobalWorkMapFactory}}
returns always same {{BaseWork}} corresponding to the given path if path is the same. 

{code}
  public Map<Path, BaseWork> get(Configuration conf) {
    if (LlapIoProxy.isDaemon()
      // For LLAP
    }

    if (gWorkMap == null) {
      gWorkMap = new HashMap<Path, BaseWork>();
    }
    // This is global object shared among multiple {{ExecDriver}}s. 
    return gWorkMap;
  }
{code}

{{MapWork}} or its HashTable should be {{synchronized}} for the case of multiple root tasks
(which cause multiple ExecDrivers).



> To sometimes deadlock when run a query
> --------------------------------------
>
>                 Key: HIVE-11749
>                 URL: https://issues.apache.org/jira/browse/HIVE-11749
>             Project: Hive
>          Issue Type: Bug
>    Affects Versions: 0.13.0
>            Reporter: Ryu Kobayashi
>         Attachments: HIVE-11749.stack-tarace.txt
>
>
> But not always, to deadlock when it run the query. Environment are as follows:
> * Hadoop 2.6.0
> * Hive 0.13
> * JDK 1.7.0_79
> It will attach the stack trace.



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

Mime
View raw message