You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@flink.apache.org by "Gyula Fora (Jira)" <ji...@apache.org> on 2023/02/24 15:02:00 UTC
[jira] [Created] (FLINK-31215) Backpropagate processing rate limits from non-scalable bottlenecks to upstream operators
Gyula Fora created FLINK-31215:
----------------------------------
Summary: Backpropagate processing rate limits from non-scalable bottlenecks to upstream operators
Key: FLINK-31215
URL: https://issues.apache.org/jira/browse/FLINK-31215
Project: Flink
Issue Type: New Feature
Components: Autoscaler, Kubernetes Operator
Reporter: Gyula Fora
The current algorithm scales operators based on input data rates by propagating it forward through the graph.
However there are cases where a certain operators processing capacity is limited either because it has a set maxParallelism or the users excludes it from scaling (or otherwise the capacity doesnt increase with scaling).
In these cases it doesn't make sense to scale upstream operators to the target data rate if the job is going to be bottlenecked by a downstream operator. But instead we should backpropagate the limit based on the non-scalable bottleneck.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)