[
https://issues.apache.org/jira/browse/SAMZA-123?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13984224#comment-13984224
]
Garry Turkington commented on SAMZA-123:
----------------------------------------
Also late to the party here but a few thoughts:
I'm with Martin/Jakob on having a liking for cohort. The reason is that I feel
the other proposed names just say "this is a task/shard/whatever". But the
creation of the thing has a degree of logic being applied and whatever term is
used I think it's useful to help reinforce that the nature of how those SSPs
were pulled together was far from arbitrary. And sorry Martin but I think shard
is completely polluted by its past in the database world.
Pluggability: I agree it's important to experiment in youth but I also fear
that if this ability is exposed that it will become something that new users
of Samza start playing with and end up getting very confused or otherwise
disillusioned with the system. I also think that the strategies proposed to be
delivered within this Jira will cover the vast vast majority of cases and its
not obvious to me just how beneficial having this as a public pluggable
capability would be.
I was also interested in the idea from Jay re using ZK but I get Chris' point
on multiple solutions. Maybe its conceptual but I feel a bit odd about pushing
increasing layers of config and state into Kafka (or indeed any other supported
system). And on that point is a reliance on KAFKA-1000 not going to add more
complexity to other SystemProducer implementations?
> Move topic partition grouping to the AM and generalize
> ------------------------------------------------------
>
> Key: SAMZA-123
> URL: https://issues.apache.org/jira/browse/SAMZA-123
> Project: Samza
> Issue Type: Sub-task
> Components: container
> Affects Versions: 0.6.0
> Reporter: Jakob Homan
> Assignee: Jakob Homan
> Attachments: SAMZA-123-design-doc.md, SAMZA-123-design-doc.pdf
>
>
> Currently the AM sends a set of all the topics and partitions to the
> container, which then groups them by partition and assigns each set to a task
> instance. By moving the grouping to the AM, we can assign arbitrary groups to
> task instances, which will allow more partitioning strategies, as discussed
> in SAMZA-71.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
