[
https://issues.apache.org/jira/browse/HAMA-906?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Aguinaldo Pontes updated HAMA-906:
----------------------------------
Description:
When implementing an algorithm on Apache Hama platform, we noticed that when
the first vertex of the split received a message, all the others became
activated (being previously halted), even if they didn't receive a message.
We think that this happens in the line 254 in GraphJobRunner, in the
doSuperstep method (
https://github.com/apache/hama/blob/trunk/graph/src/main/java/org/apache/hama/graph/GraphJobRunner.java#L254).
We think that in this line it misses the following verification:
if(currentMessage!=null && ... )
Could you confirm?
Thanks.
was:
When implementing an algorithm on Apache Hama platform, we noticed that when
the first vertex of the split received a message, all the others became
activated (being previously halted), even if they didn't receive a message.
We think that this happens in the line 254 in GraphJobRunner, in the
doSuperstep method (
https://github.com/apache/hama/blob/trunk/graph/src/main/java/org/apache/hama/graph/GraphJobRunner.java#L254).
We think that in this line it misses the following verification:
if(currentMessage!=null && ... )
Are we right?
> Automatic activation of halted vertices without received messages
> -----------------------------------------------------------------
>
> Key: HAMA-906
> URL: https://issues.apache.org/jira/browse/HAMA-906
> Project: Hama
> Issue Type: Bug
> Components: graph
> Affects Versions: 0.6.4
> Reporter: Aguinaldo Pontes
> Labels: easyfix
>
> When implementing an algorithm on Apache Hama platform, we noticed that when
> the first vertex of the split received a message, all the others became
> activated (being previously halted), even if they didn't receive a message.
> We think that this happens in the line 254 in GraphJobRunner, in the
> doSuperstep method (
> https://github.com/apache/hama/blob/trunk/graph/src/main/java/org/apache/hama/graph/GraphJobRunner.java#L254).
> We think that in this line it misses the following verification:
> if(currentMessage!=null && ... )
> Could you confirm?
> Thanks.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
