[
https://issues.apache.org/jira/browse/FLINK-34452?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17819086#comment-17819086
]
Weijie Guo commented on FLINK-34452:
------------------------------------
I have tested this feature on a yarn cluster:
1. Set {{taskmanager.numberOfTaskSlots=2}},
{{slotmanager.number-of-slots.min=7}}, and {{slot.idle.timeout=50000}}.
2. Start a yarn session cluster.
3. From ActiveResourceManager's log that 4 workers (TM) were requested and
successfully registered.
4. These workers are not be destroyed even after idle timeout.
Everything looks as expected then.
> Release Testing: Verify FLINK-15959 Add min number of slots configuration to
> limit total number of slots
> --------------------------------------------------------------------------------------------------------
>
> Key: FLINK-34452
> URL: https://issues.apache.org/jira/browse/FLINK-34452
> Project: Flink
> Issue Type: Sub-task
> Components: Runtime / Coordination
> Affects Versions: 1.19.0
> Reporter: xiangyu feng
> Assignee: Weijie Guo
> Priority: Blocker
> Labels: release-testing
> Fix For: 1.19.0
>
>
> Test Suggestion:
> # Prepare configuration options:
> ** taskmanager.numberOfTaskSlots = 2,
> ** slotmanager.number-of-slots.min = 7,
> ** slot.idle.timeout = 50000
> # Setup a Flink session Cluster on Yarn or Native Kubernetes based on
> following docs:
> **
> [https://nightlies.apache.org/flink/flink-docs-master/docs/deployment/resource-providers/yarn/#starting-a-flink-session-on-yarn]
> **
> [https://nightlies.apache.org/flink/flink-docs-master/docs/deployment/resource-providers/native_kubernetes/#starting-a-flink-session-on-kubernetes]
> # Verify that 4 TaskManagers will be registered even though no jobs has been
> submitted
> # Verify that these TaskManagers will not be destroyed after 50 seconds.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
