[ https://issues.apache.org/jira/browse/FLINK-26356?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17497937#comment-17497937 ]
Yang Wang commented on FLINK-26356: ----------------------------------- I think we could always use the {{<cluster-id>-rest.<namespace>}} to access the Flink rest server no matter the HA is enabled or not. Right? > Revisit the create of RestClusterClient > --------------------------------------- > > Key: FLINK-26356 > URL: https://issues.apache.org/jira/browse/FLINK-26356 > Project: Flink > Issue Type: Sub-task > Components: Kubernetes Operator > Reporter: Aitozi > Priority: Major > > The clusterClient is built as below. The config is mixed up with the > FlinkDeploymentSpec and local default config. > {code:java} > final int port = config.getInteger(RestOptions.PORT); > final String host = > config.getString( > RestOptions.ADDRESS, String.format("%s-rest.%s", clusterId, > namespace)); > final String restServerAddress = String.format("http://%s:%s", host, port); > {code} > But the {{RestOptions.ADDRESS}} is generated at the entrypoint when the HA is > enabled, so the option can not obtain from the FlinkDeploymentSpec. > Furthermore, the default rest url is not suitable for all the service type. I > think we should extract the rest endpoint from the Flink external service. -- This message was sent by Atlassian Jira (v8.20.1#820001)