[ 
https://issues.apache.org/jira/browse/GEODE-5884?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16719721#comment-16719721
 ] 

ASF subversion and git services commented on GEODE-5884:
--------------------------------------------------------

Commit e54e33d19135af64177145d96799d9974812d45d in geode's branch 
refs/heads/master from [~huynhja]
[ https://gitbox.apache.org/repos/asf?p=geode.git;h=e54e33d ]

GEODE-5884: Added new command and restored pre 1.8 region function behavior 
(#2829)

* Due to the refactor that caused us to lose wrapping of exceptions,
  the fix to GEODE-5884 cannot be applied to GEODE 1.0-1.7 clients
* Adding new command class will allow clients to identify which function class
  it is expecting to execute and what behavior to expect


> Executing a function with non HA should return any available partial results
> ----------------------------------------------------------------------------
>
>                 Key: GEODE-5884
>                 URL: https://issues.apache.org/jira/browse/GEODE-5884
>             Project: Geode
>          Issue Type: Bug
>          Components: functions
>    Affects Versions: 1.0.0-incubating
>            Reporter: Jason Huynh
>            Assignee: Jason Huynh
>            Priority: Major
>              Labels: pull-request-available
>          Time Spent: 3h
>  Remaining Estimate: 0h
>
> When a function is executed without single hop, it is sent to a 
> "coordinating" node.  This node then creates tasks that are executed on other 
> remote nodes and itself.  If any of the remote nodes or the local node gets 
> an exception, it will accumulate the exception and any results it has 
> currently received.  It then completes and sends those results back to the 
> client.
> There is a flag to waitOnException.  We should set this to true if the 
> function is non HA.  This will have the coordinating node wait for all 
> results before sending back to the client.  The work is going to get executed 
> on the remote nodes anyways, so why not wait for the results?
> This will mirror a single-hop execution where a single node may fail.  If it 
> does, the client still receives results from the other nodes.
>  



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to