You are viewing a plain text version of this content. The canonical link for it is here.
Posted to common-dev@hadoop.apache.org by "Runping Qi (JIRA)" <ji...@apache.org> on 2008/08/27 03:14:44 UTC

[jira] Created: (HADOOP-4033) Task tracker should not ask for tasks if its temp disk space is almost full

Task tracker should not ask for tasks if its temp disk space is almost full
---------------------------------------------------------------------------

                 Key: HADOOP-4033
                 URL: https://issues.apache.org/jira/browse/HADOOP-4033
             Project: Hadoop Core
          Issue Type: Bug
    Affects Versions: 0.17.1
            Reporter: Runping Qi



I observed a case where a task tracker still asked for task even though the available disk space on the machine is less than 1%.
Consequently, it had hard time to finish the tasks.


-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Commented: (HADOOP-4033) Task tracker should not ask for tasks if its temp disk space is almost full

Posted by "Sameer Paranjpye (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/HADOOP-4033?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12668789#action_12668789 ] 

Sameer Paranjpye commented on HADOOP-4033:
------------------------------------------

It should also be possible to set minspacestart to a reasonable value in mapred-site.xml, no? 

> Task tracker should not ask for tasks if its temp disk space is almost full
> ---------------------------------------------------------------------------
>
>                 Key: HADOOP-4033
>                 URL: https://issues.apache.org/jira/browse/HADOOP-4033
>             Project: Hadoop Core
>          Issue Type: Bug
>    Affects Versions: 0.17.1
>            Reporter: Runping Qi
>            Assignee: Ravi Gummadi
>
> I observed a case where a task tracker still asked for task even though the available disk space on the machine is less than 1%.
> Consequently, it had hard time to finish the tasks.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Commented: (HADOOP-4033) Task tracker should not ask for tasks if its temp disk space is almost full

Posted by "Ravi Gummadi (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/HADOOP-4033?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12668795#action_12668795 ] 

Ravi Gummadi commented on HADOOP-4033:
--------------------------------------

Devaraj suggested porting JIRA-657's patch to branch 18 would be better. It applies fine to branch 18. Looking into the code of that patch to understand what is happening.

> Task tracker should not ask for tasks if its temp disk space is almost full
> ---------------------------------------------------------------------------
>
>                 Key: HADOOP-4033
>                 URL: https://issues.apache.org/jira/browse/HADOOP-4033
>             Project: Hadoop Core
>          Issue Type: Bug
>    Affects Versions: 0.17.1
>            Reporter: Runping Qi
>            Assignee: Ravi Gummadi
>
> I observed a case where a task tracker still asked for task even though the available disk space on the machine is less than 1%.
> Consequently, it had hard time to finish the tasks.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Commented: (HADOOP-4033) Task tracker should not ask for tasks if its temp disk space is almost full

Posted by "Runping Qi (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/HADOOP-4033?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12625939#action_12625939 ] 

Runping Qi commented on HADOOP-4033:
------------------------------------


In the case where multiple  temp dirs are configured, hadoop should avoid using a dir when it's available space is below certain threshold.


> Task tracker should not ask for tasks if its temp disk space is almost full
> ---------------------------------------------------------------------------
>
>                 Key: HADOOP-4033
>                 URL: https://issues.apache.org/jira/browse/HADOOP-4033
>             Project: Hadoop Core
>          Issue Type: Bug
>    Affects Versions: 0.17.1
>            Reporter: Runping Qi
>
> I observed a case where a task tracker still asked for task even though the available disk space on the machine is less than 1%.
> Consequently, it had hard time to finish the tasks.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Commented: (HADOOP-4033) Task tracker should not ask for tasks if its temp disk space is almost full

Posted by "Ravi Gummadi (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/HADOOP-4033?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12665787#action_12665787 ] 

Ravi Gummadi commented on HADOOP-4033:
--------------------------------------

Configuring mapred.local.dir.minspacestart by user should solve this problem.

Currently, the default value of mapred.local.dir.minspacestart is 0 bytes. How about changing this default value to a +ve value(say 1MB) in mapred.default.xml ?

Thoughts ?

> Task tracker should not ask for tasks if its temp disk space is almost full
> ---------------------------------------------------------------------------
>
>                 Key: HADOOP-4033
>                 URL: https://issues.apache.org/jira/browse/HADOOP-4033
>             Project: Hadoop Core
>          Issue Type: Bug
>    Affects Versions: 0.17.1
>            Reporter: Runping Qi
>            Assignee: Ravi Gummadi
>
> I observed a case where a task tracker still asked for task even though the available disk space on the machine is less than 1%.
> Consequently, it had hard time to finish the tasks.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Commented: (HADOOP-4033) Task tracker should not ask for tasks if its temp disk space is almost full

Posted by "Arun C Murthy (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/HADOOP-4033?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12668964#action_12668964 ] 

Arun C Murthy commented on HADOOP-4033:
---------------------------------------

bq. Actually, I'd propose that we backport the 0.19 items - HADOOP-657 (and probably another related issue - HADOOP-9) to handle this issue.

Uh, we can't add new features to a released branch?

> Task tracker should not ask for tasks if its temp disk space is almost full
> ---------------------------------------------------------------------------
>
>                 Key: HADOOP-4033
>                 URL: https://issues.apache.org/jira/browse/HADOOP-4033
>             Project: Hadoop Core
>          Issue Type: Bug
>    Affects Versions: 0.17.1
>            Reporter: Runping Qi
>            Assignee: Ravi Gummadi
>
> I observed a case where a task tracker still asked for task even though the available disk space on the machine is less than 1%.
> Consequently, it had hard time to finish the tasks.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Assigned: (HADOOP-4033) Task tracker should not ask for tasks if its temp disk space is almost full

Posted by "Ravi Gummadi (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/HADOOP-4033?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Ravi Gummadi reassigned HADOOP-4033:
------------------------------------

    Assignee: Ravi Gummadi

> Task tracker should not ask for tasks if its temp disk space is almost full
> ---------------------------------------------------------------------------
>
>                 Key: HADOOP-4033
>                 URL: https://issues.apache.org/jira/browse/HADOOP-4033
>             Project: Hadoop Core
>          Issue Type: Bug
>    Affects Versions: 0.17.1
>            Reporter: Runping Qi
>            Assignee: Ravi Gummadi
>
> I observed a case where a task tracker still asked for task even though the available disk space on the machine is less than 1%.
> Consequently, it had hard time to finish the tasks.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Commented: (HADOOP-4033) Task tracker should not ask for tasks if its temp disk space is almost full

Posted by "Devaraj Das (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/HADOOP-4033?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12668791#action_12668791 ] 

Devaraj Das commented on HADOOP-4033:
-------------------------------------

Actually, I'd propose that we backport the 0.19 items - HADOOP-657 (and probably another related issue - HADOOP-9) to handle this issue.. HADOOP-657 seems like a better approach for the problem described.. thoughts?

> Task tracker should not ask for tasks if its temp disk space is almost full
> ---------------------------------------------------------------------------
>
>                 Key: HADOOP-4033
>                 URL: https://issues.apache.org/jira/browse/HADOOP-4033
>             Project: Hadoop Core
>          Issue Type: Bug
>    Affects Versions: 0.17.1
>            Reporter: Runping Qi
>            Assignee: Ravi Gummadi
>
> I observed a case where a task tracker still asked for task even though the available disk space on the machine is less than 1%.
> Consequently, it had hard time to finish the tasks.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Commented: (HADOOP-4033) Task tracker should not ask for tasks if its temp disk space is almost full

Posted by "Devaraj Das (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/HADOOP-4033?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12668793#action_12668793 ] 

Devaraj Das commented on HADOOP-4033:
-------------------------------------

In addition to HADOOP-657, setting some sensible value for the minspacestart in mapred-site.xml should also be done

> Task tracker should not ask for tasks if its temp disk space is almost full
> ---------------------------------------------------------------------------
>
>                 Key: HADOOP-4033
>                 URL: https://issues.apache.org/jira/browse/HADOOP-4033
>             Project: Hadoop Core
>          Issue Type: Bug
>    Affects Versions: 0.17.1
>            Reporter: Runping Qi
>            Assignee: Ravi Gummadi
>
> I observed a case where a task tracker still asked for task even though the available disk space on the machine is less than 1%.
> Consequently, it had hard time to finish the tasks.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.