[
https://issues.apache.org/jira/browse/FLINK-18109?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17125740#comment-17125740
]
Yangze Guo edited comment on FLINK-18109 at 6/4/20, 9:40 AM:
-------------------------------------------------------------
Test log:
* Standalone mode, Linux, 2 NVIDIA GPUs
** normal mode, works well.
** coordination mode, works well.
* Kubernetes mode
** build a custom image from Flink-1.11.0-RC1
** {{bin/kubernetes-session.sh and submit WordCount job}}
** log of task executor shows everything works well.
* Yarn
** 2.10, works well.
** 3.1, works well.
was (Author: karmagyz):
Test log:
* Standalone mode, Linux, 2 NVIDIA GPUs
** normal mode, works well.
** coordination mode, works well.
* Kubernetes mode
** build a custom image from Flink-1.11.0-RC1
** {{bin/kubernetes-session.sh and submit WordCount job}}
** log of task executor shows everything works well
* Yarn
** Todo
> Manually test external resource framework with GPUDriver
> --------------------------------------------------------
>
> Key: FLINK-18109
> URL: https://issues.apache.org/jira/browse/FLINK-18109
> Project: Flink
> Issue Type: Sub-task
> Affects Versions: 1.11.0
> Reporter: Yangze Guo
> Priority: Blocker
> Fix For: 1.11.0
>
>
> We need to:
> - Test it in standalone mode.
> ** The discovery script works well in default mode
> ** The discovery script works well in coordination mode.
> - Test it with Yarn 2.10 and 3.1.
> ** GPU resources are allocated successfully.
> - Test it with Kubernetes
> ** GPU resources are allocated successfully.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
