[
https://issues.apache.org/jira/browse/FLINK-40866?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Dale Lane updated FLINK-40866:
------------------------------
Description:
Wires up the {{FlinkMiniCluster}} components added in the preceding sub-tasks
into a working controller.
Adds {{FlinkMiniClusterController}} (JOSDK {{{}Reconciler{}}}),
{{{}FlinkMiniClusterContext{}}}, the informer event source on the MiniCluster
Deployment, status and event recording for the new kind, and registration in
{{FlinkOperator}} including canary-based health probing.
With this change a {{FlinkMiniCluster}} resource will be reconciled end to end:
the operator deploys the pod, observes it, reports job and pod status, and
cleans up on deletion.
A FlinkMiniCluster may have the same name as a FlinkDeployment in the same
namespace. They are different kinds, so no cross-kind name validation is
planned. As with a FlinkDeployment, the Kubernetes objects that the controller
creates may collide with objects that already exist, and this should be handled
the same way: failing on collision, and ownership-aware cleanup.
An end-to-end test covering the basic lifecycle should be included.
_(Note: {{FlinkResourceListener}} has no {{FlinkMiniCluster}} callbacks, so
listener notification and audit logging are not wired for this kind.)_
was:
Wires up the {{FlinkMiniCluster}} components added in the preceding sub-tasks
into a working controller.
Adds {{FlinkMiniClusterController}} (JOSDK {{{}Reconciler{}}}),
{{{}FlinkMiniClusterContext{}}}, the informer event source on the MiniCluster
Deployment, status and event recording for the new kind, and registration in
{{FlinkOperator}} including canary-based health probing.
With this change a {{FlinkMiniCluster}} resource will be reconciled end to end:
the operator deploys the pod, observes it, reports job and pod status, and
cleans up on deletion.
An end-to-end test covering the basic lifecycle should be included.
_(Note: {{FlinkResourceListener}} has no {{FlinkMiniCluster}} callbacks, so
listener notification and audit logging are not wired for this kind.)_
> Add FlinkMiniClusterController and register the FlinkMiniCluster kind
> ---------------------------------------------------------------------
>
> Key: FLINK-40866
> URL: https://issues.apache.org/jira/browse/FLINK-40866
> Project: Flink
> Issue Type: Sub-task
> Components: Kubernetes Operator
> Reporter: Dale Lane
> Priority: Major
>
> Wires up the {{FlinkMiniCluster}} components added in the preceding sub-tasks
> into a working controller.
> Adds {{FlinkMiniClusterController}} (JOSDK {{{}Reconciler{}}}),
> {{{}FlinkMiniClusterContext{}}}, the informer event source on the MiniCluster
> Deployment, status and event recording for the new kind, and registration in
> {{FlinkOperator}} including canary-based health probing.
> With this change a {{FlinkMiniCluster}} resource will be reconciled end to
> end: the operator deploys the pod, observes it, reports job and pod status,
> and cleans up on deletion.
> A FlinkMiniCluster may have the same name as a FlinkDeployment in the same
> namespace. They are different kinds, so no cross-kind name validation is
> planned. As with a FlinkDeployment, the Kubernetes objects that the
> controller creates may collide with objects that already exist, and this
> should be handled the same way: failing on collision, and ownership-aware
> cleanup.
> An end-to-end test covering the basic lifecycle should be included.
> _(Note: {{FlinkResourceListener}} has no {{FlinkMiniCluster}} callbacks, so
> listener notification and audit logging are not wired for this kind.)_
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
