[
https://issues.apache.org/jira/browse/FLINK-20411?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17240443#comment-17240443
]
Yang Wang commented on FLINK-20411:
-----------------------------------
I think the ConfigOption in Map type could support complex structure now. All
the thing you need to do is escape the strings. You could find more information
here[1]. In your case, the following config option could work.
{code:java}
-Dkubernetes.jobmanager.annotations="'k1:''{"a1":"v1","a2":"v2"}''','k2:''{"a1":"v1","a2":"v2"}'''"{code}
[1].
https://github.com/apache/flink/blob/5af1d007e26e0fb437028d5882d79fe09baf937a/flink-core/src/main/java/org/apache/flink/configuration/StructuredOptionsSplitter.java#L56
> The annotation config do not support complex structure
> ------------------------------------------------------
>
> Key: FLINK-20411
> URL: https://issues.apache.org/jira/browse/FLINK-20411
> Project: Flink
> Issue Type: Improvement
> Components: Deployment / Kubernetes
> Affects Versions: 1.11.2
> Reporter: Aitozi
> Priority: Minor
>
> Now we support user to set annotations by these config, the format is
> "k1:v1,k2:v2".
> # kubernetes.rest-service.annotations
> # kubernetes.jobmanager.annotations
> # kubernetes.taskmanager.annotations
> But the annotation may be more complex structure like
> "k1:\{"a1":"v1","a2":"v2"},k2:\{"a1":"v1","a2":"v2"}", It can not be parsed.
> Shall we support annotation and labels just like the environment variables
> use a prefix split mechanism, like :
> # kubernetes.rest-service.annotation.k1="\{"a1":"v1","a2":"v2"}"
> By this we can set some default cluster annotations and labels and no need to
> worry about overridden by user config.
>
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
