You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@flink.apache.org by "Rui Fan (Jira)" <ji...@apache.org> on 2024/02/02 05:56:00 UTC

[jira] [Commented] (FLINK-34178) The ScalingTracking of autoscaler is wrong

    [ https://issues.apache.org/jira/browse/FLINK-34178?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17813488#comment-17813488 ] 

Rui Fan commented on FLINK-34178:
---------------------------------

Merged to main(1.8.0) via:
 * 2bd6b1b38171ce821509d46d687291b876f72a2e
 * 746a996732e56e573f7882764c82e2faa2cba71d

> The ScalingTracking of autoscaler is wrong
> ------------------------------------------
>
>                 Key: FLINK-34178
>                 URL: https://issues.apache.org/jira/browse/FLINK-34178
>             Project: Flink
>          Issue Type: Bug
>          Components: Autoscaler
>    Affects Versions: kubernetes-operator-1.8.0
>            Reporter: Rui Fan
>            Assignee: Rui Fan
>            Priority: Major
>              Labels: pull-request-available
>
> The ScalingTracking of autoscaler is wrong, it's always greater than AutoScalerOptions#STABILIZATION_INTERVAL.
> h2. Reason:
> When flink job isStabilizing, ScalingMetricCollector#updateMetrics will return a empty metric history. In the JobAutoScalerImpl#runScalingLogic method, if `collectedMetrics.getMetricHistory().isEmpty()` , we don't update the ScalingTracking.
>  
> The default value of AutoScalerOptions#STABILIZATION_INTERVAL is 5 min, so the restartTime is always greater than 5 min.
> However, it's quick when we use rescale api.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)