[
https://issues.apache.org/jira/browse/FLINK-10407?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16864753#comment-16864753
]
Flavio Junqueira commented on FLINK-10407:
------------------------------------------
[~till.rohrmann] I was giving a look at the design document posted in this
issue and I have a couple of questions. Is it ok if I post them here or is it
better to do it on the sub-tasks or even the document itself?
1- According to the proposal, each operator can have its own rescaling policy,
and it makes sense. I'm wondering about constraints across operators, though.
Say that I give the target to an operator, but only the minimum to one or more
operators downstream. Can it cause any runtime instability? If so, then perhaps
we need also a way to specify constraints across operators. I'm interested in
your feedback here.
2- It might be early to ask about the rescaling policy as there is little
information about it in the document, but I'm interested in your current
thoughts on how the interface is going to look like and how implementations are
going to be used with a job. Specifically, I'm interested in how to connect the
auto-scaling of [Pravega|http://github.com/pravega/pravega] to the Flink
dynamic scaling so that we have scaling end-to-end.
> Reactive container mode
> -----------------------
>
> Key: FLINK-10407
> URL: https://issues.apache.org/jira/browse/FLINK-10407
> Project: Flink
> Issue Type: New Feature
> Components: Runtime / Coordination
> Affects Versions: 1.7.0
> Reporter: Till Rohrmann
> Assignee: Till Rohrmann
> Priority: Major
>
> The reactive container mode is a new operation mode where a Flink cluster
> will react to newly available resources (e.g. started by an external service)
> and make use of them by rescaling the existing job.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
