[jira] [Commented] (SAMZA-152) Hello-samza should show how to use a store

["Martin Kleppmann (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SAMZA-152?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13923047#comment-13923047 ] 

Martin Kleppmann commented on SAMZA-152:
----------------------------------------

I've opened a new pull request on top of master: https://github.com/linkedin/hello-samza/pull/33 — tested together with the patch for SAMZA-169, it seems to be working nicely.

Regarding breaking the aggregate count: rather than setting num.partitions=1 in the Kafka config, how about we make the wikipedia-parser job output messages with a constant key, so that they all go into the same partition? That way we can still demonstrate parallelism in the wikipedia-parser job. I've added a separate commit to the pull request, and it seems quite straightforward.

But if you prefer that we just set num.partitions=1 in bin/grid, that's fine with me too — I can update the PR if you prefer that way.

> Hello-samza should show how to use a store
> ------------------------------------------
>
>                 Key: SAMZA-152
>                 URL: https://issues.apache.org/jira/browse/SAMZA-152
>             Project: Samza
>          Issue Type: Improvement
>            Reporter: Martin Kleppmann
>            Assignee: Martin Kleppmann
>
> Given that state management is one of Samza's headline features, it would be good if hello-samza included a demo of how to use it. The wikipedia-stats job would be suitable, since it already performs an in-memory aggregation. We could change it to do a persistent aggregate instead.



--
This message was sent by Atlassian JIRA
(v6.2#6252)