[
https://issues.apache.org/jira/browse/CRUNCH-611?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Micah Whitacre updated CRUNCH-611:
----------------------------------
Attachment: CRUNCH-611.patch
So this patch provides a basic API for reading/writing Kafka offsets. It then
also provides a simple implementation that reads/writes the values from HDFS.
In theory this then should make regularly schedule Crunch pipeline's easier to
implement with regard to offset management.
I did add a few optional dependencies so hopefully these won't cause too bad of
conflicts with the Hadoop stack. We aren't having a problem on our cluster but
didn't universally check. We are also setting out classpath first and running
through Oozie so that changes classpath ordering as well.
> Simplified Kafka Offset Management in HDFS
> ------------------------------------------
>
> Key: CRUNCH-611
> URL: https://issues.apache.org/jira/browse/CRUNCH-611
> Project: Crunch
> Issue Type: Improvement
> Components: IO
> Reporter: Micah Whitacre
> Assignee: Micah Whitacre
> Attachments: CRUNCH-611.patch
>
>
> With the KafkaSource the responsibility of offset management is the burden of
> the consumer. With some simple APIs it is actually trivial to support
> read/storing these offsets in an HDFS directory as checkpoints for the source.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
