You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@giraph.apache.org by "Maja Kabiljo (JIRA)" <ji...@apache.org> on 2012/10/02 02:44:07 UTC

[jira] [Created] (GIRAPH-355) Partition.readFields crashes

Maja Kabiljo created GIRAPH-355:
-----------------------------------

             Summary: Partition.readFields crashes
                 Key: GIRAPH-355
                 URL: https://issues.apache.org/jira/browse/GIRAPH-355
             Project: Giraph
          Issue Type: Bug
            Reporter: Maja Kabiljo


Partition:157 (vertex.getContext().progress();) will crash since graphState for vertex is null at this point. We need to get Context inside Partition in order to be able to call progress here. 
The reason why this wasn't noticed is that our local tests don't try checkpoint restart at the moment, and that's the only place where Partition.readFields is called.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Commented] (GIRAPH-355) Partition.readFields crashes

Posted by "Avery Ching (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/GIRAPH-355?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13468249#comment-13468249 ] 

Avery Ching commented on GIRAPH-355:
------------------------------------

+1, committing.
                
> Partition.readFields crashes
> ----------------------------
>
>                 Key: GIRAPH-355
>                 URL: https://issues.apache.org/jira/browse/GIRAPH-355
>             Project: Giraph
>          Issue Type: Bug
>            Reporter: Maja Kabiljo
>            Assignee: Maja Kabiljo
>         Attachments: GIRAPH-355.patch
>
>
> Partition:157 (vertex.getContext().progress();) will crash since graphState for vertex is null at this point. We need to get Context inside Partition in order to be able to call progress here. 
> The reason why this wasn't noticed is that our local tests don't try checkpoint restart at the moment, and that's the only place where Partition.readFields is called.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Assigned] (GIRAPH-355) Partition.readFields crashes

Posted by "Maja Kabiljo (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/GIRAPH-355?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Maja Kabiljo reassigned GIRAPH-355:
-----------------------------------

    Assignee: Maja Kabiljo
    
> Partition.readFields crashes
> ----------------------------
>
>                 Key: GIRAPH-355
>                 URL: https://issues.apache.org/jira/browse/GIRAPH-355
>             Project: Giraph
>          Issue Type: Bug
>            Reporter: Maja Kabiljo
>            Assignee: Maja Kabiljo
>         Attachments: GIRAPH-355.patch
>
>
> Partition:157 (vertex.getContext().progress();) will crash since graphState for vertex is null at this point. We need to get Context inside Partition in order to be able to call progress here. 
> The reason why this wasn't noticed is that our local tests don't try checkpoint restart at the moment, and that's the only place where Partition.readFields is called.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Commented] (GIRAPH-355) Partition.readFields crashes

Posted by "Avery Ching (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/GIRAPH-355?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13467497#comment-13467497 ] 

Avery Ching commented on GIRAPH-355:
------------------------------------

This is fixable if Partition takes a Progressable input and calls progress from there rather than using the vertex progress.  Do you have a failure case that we can look at? 
                
> Partition.readFields crashes
> ----------------------------
>
>                 Key: GIRAPH-355
>                 URL: https://issues.apache.org/jira/browse/GIRAPH-355
>             Project: Giraph
>          Issue Type: Bug
>            Reporter: Maja Kabiljo
>
> Partition:157 (vertex.getContext().progress();) will crash since graphState for vertex is null at this point. We need to get Context inside Partition in order to be able to call progress here. 
> The reason why this wasn't noticed is that our local tests don't try checkpoint restart at the moment, and that's the only place where Partition.readFields is called.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Updated] (GIRAPH-355) Partition.readFields crashes

Posted by "Maja Kabiljo (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/GIRAPH-355?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Maja Kabiljo updated GIRAPH-355:
--------------------------------

    Attachment: GIRAPH-355.patch

Here is the patch, there were quite a few places where Context needed to be added so we would always have it for Partition.

Crash happens when you run TestAutoCheckpoint in pseudo distributed mode. Or you can add one line to TestManualCheckpoint to set the superstep which we want to restart from:
restartedJob.getConfiguration().setLong(GiraphConfiguration.RESTART_SUPERSTEP, 4);
This test will end up crashing even when you remove progress calls because GIRAPH-293 is not in yet, but this null exception happens first.
                
> Partition.readFields crashes
> ----------------------------
>
>                 Key: GIRAPH-355
>                 URL: https://issues.apache.org/jira/browse/GIRAPH-355
>             Project: Giraph
>          Issue Type: Bug
>            Reporter: Maja Kabiljo
>            Assignee: Maja Kabiljo
>         Attachments: GIRAPH-355.patch
>
>
> Partition:157 (vertex.getContext().progress();) will crash since graphState for vertex is null at this point. We need to get Context inside Partition in order to be able to call progress here. 
> The reason why this wasn't noticed is that our local tests don't try checkpoint restart at the moment, and that's the only place where Partition.readFields is called.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira