[
https://issues.apache.org/jira/browse/FLINK-22802?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17379714#comment-17379714
]
Chesnay Schepler commented on FLINK-22802:
------------------------------------------
In our e2e tests we use minikube 1.8.2, which should rely on [Kubernetes
1.17.3|https://github.com/kubernetes/minikube/blob/v1.8.2/go.sum#L1055] .
I suppose we will eventually need to switch to 5.X in any case.
[~fly_in_gis] Do you have a rough idea on the effort required for migrating to
5.x?
> Bump Fabric8 Kubernetes Client to >= 5.X
> ----------------------------------------
>
> Key: FLINK-22802
> URL: https://issues.apache.org/jira/browse/FLINK-22802
> Project: Flink
> Issue Type: Improvement
> Components: Deployment / Kubernetes
> Affects Versions: 1.13.0, 1.13.1, 1.12.4
> Environment: Kubernetes 1.19/1.20 provided as part of Openshift
> 4.6/4.7
> Flink 1.13.0 and 1.13.1-rc1
> Reporter: Enrique Lacal
> Priority: Minor
> Labels: auto-deprioritized-major
>
> Hi All,
> Currently, Flink is using version 4.9.2 of the Fabric8 Kubernetes Client
> which does not support new versions of the Kubernetes API such as 1.19 or
> 1.20 as pointed out by their [Compatibility
> Matrix|https://github.com/fabric8io/kubernetes-client#compatibility-matrix]
> which is support in 5.4.0 or above.
> As far as I have seen in the Flink documentation, Flink supports `Kubernetes
> >= 1.9.` but due to this dependency, it might not be the case. Is there a
> plan to update this dependency?
> What is the plan moving forwards when new versions of Kubernetes are released?
> I am raising this because I have been testing Flink HA Session Cluster on
> Kubernetes 1.19 and 1.20 and I have encountered some frequent errors that
> force the JM pods to restart or even result in an unrecoverable state.
> Thanks!
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
