[ 
https://issues.apache.org/jira/browse/HBASE-15320?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16054500#comment-16054500
 ] 

Mike Wingert commented on HBASE-15320:
--------------------------------------

Thanks for the feedback!

> Please start patches against the master branch. We can figure out what 
> release lines a new feature can get into once things get hammered out more.

Sorry about that, I'll re-work the patch.  

> What do we gain by running edits through the proxy servers rather than just 
> putting directly into the kafka broker directly from the replication system?

I wanted to isolate the kafka logic out of the region server completely, in 
case something happened on the kafka side.  I also wanted to use the 
replication peer system as a way to buffer things in case kafka slows down.

> How about some docs?

Docs are on the way =-)

> HBase connector for Kafka Connect
> ---------------------------------
>
>                 Key: HBASE-15320
>                 URL: https://issues.apache.org/jira/browse/HBASE-15320
>             Project: HBase
>          Issue Type: New Feature
>          Components: Replication
>            Reporter: Andrew Purtell
>            Assignee: Mike Wingert
>              Labels: beginner
>             Fix For: 3.0.0, 2.1.0
>
>         Attachments: HBASE-15320.txt
>
>
> Implement an HBase connector with source and sink tasks for the Connect 
> framework (http://docs.confluent.io/2.0.0/connect/index.html) available in 
> Kafka 0.9 and later.
> See also: 
> http://www.confluent.io/blog/announcing-kafka-connect-building-large-scale-low-latency-data-pipelines
> An HBase source 
> (http://docs.confluent.io/2.0.0/connect/devguide.html#task-example-source-task)
>  could be implemented as a replication endpoint or WALObserver, publishing 
> cluster wide change streams from the WAL to one or more topics, with 
> configurable mapping and partitioning of table changes to topics.  
> An HBase sink task 
> (http://docs.confluent.io/2.0.0/connect/devguide.html#sink-tasks) would 
> persist, with optional transformation (JSON? Avro?, map fields to native 
> schema?), Kafka SinkRecords into HBase tables.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

Reply via email to