[
https://issues.apache.org/jira/browse/FLINK-32120?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Maximilian Michels updated FLINK-32120:
---------------------------------------
Fix Version/s: kubernetes-operator-1.9.0
(was: kubernetes-operator-1.8.0)
> Add autoscaler config option to disable parallelism key group alignment
> -----------------------------------------------------------------------
>
> Key: FLINK-32120
> URL: https://issues.apache.org/jira/browse/FLINK-32120
> Project: Flink
> Issue Type: Improvement
> Components: Autoscaler, Kubernetes Operator
> Reporter: Maximilian Michels
> Assignee: Maximilian Michels
> Priority: Major
> Fix For: kubernetes-operator-1.9.0
>
>
> After choosing the target parallelism for a vertex, we choose a higher
> parallelism if that parallelism leads to evenly spreading the number of key
> groups. The number of key groups is derived from the max parallelism.
> The amount of actual skew we would introduce if we did not do the alignment
> would usually be pretty low. In fact, the data itself can have an uneven load
> distribution across the keys (hot keys). In this case, the key group
> alignment is not effective.
> For experiments, we should allow disabling the key group alignment via a
> configuration option.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
