[ 
https://issues.apache.org/jira/browse/FLUME-2298?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13874911#comment-13874911
 ] 

Ted Malaska commented on FLUME-2298:
------------------------------------

Thanks Hari, Jayant, and Ashish,

I will have to update the document to reflex the following:

1. (Hari, Jayant) We can't use existing channels because we need the master and 
slave to be complete ordered and existing channels don't do that.
2. (Jayant) I need to clarify where the slave will start when the master dies.  
I would figure that the slave is in sync with the master to it will ask the 
same as it would with the next take.  Just as if it was taken from the master 
or if it was taken from a sink.  This is in the current design but it was not 
defined correctly in the doc.
3. (Jayant) I need to clarify that the slave will only discard event yet to be 
committed to a sink.  This is in the current design but it was not defined 
correctly in the doc.
4. (Ashish) Yes the option for DC_foo to continue to work without a slave 
should be configurable.  I will update the doc.
5. (Ashish) Good point I would going to use heart beats but watcher would be 
nice.  I will update the doc.
6. (Ashish) I was thinking of going more lite weight then a source and sink.  I 
was thinking a simple netty server and client.  Let me know what you think.



> Distributed Channel
> -------------------
>
>                 Key: FLUME-2298
>                 URL: https://issues.apache.org/jira/browse/FLUME-2298
>             Project: Flume
>          Issue Type: New Feature
>          Components: Channel
>            Reporter: Ted Malaska
>            Assignee: Ted Malaska
>         Attachments: FlumeDistributedChannelDesign.0.1.pdf
>
>
> This channel will allow for events to be persisted with a plugable method on 
> more then one agent or node.  
> The goal is to gain the following benefits:
> 1. Events will continue to flow to sinks with out loss or with out large 
> delay even in the case of node failure.
> 2. Protect against loss in the case of a complete single node failure



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)

Reply via email to