[GitHub] [flink] tillrohrmann edited a comment on pull request #14692: [FLINK-20944][k8s] Do not resolve the rest endpoint address when creating the RestClusterClient

[GitBox <gi...@apache.org>]

tillrohrmann edited a comment on pull request #14692:
URL: https://github.com/apache/flink/pull/14692#issuecomment-763643704


   > * We still create the RestClusterClient when deploying application/session and retrieving. For deploying, it is only used to print the web interface URL. For retrieving, it will be used to talk to the cluster.
   > * Based on above, we will use NO_ADDRESS_RESOLUTION for deploying and TRY_ADDRESS_RESOLUTION for retrieving.
   
   I don't think that this would work w/o other changes because when you deploy an attached Yarn per-job cluster, then the client will be used to query the job result which requires access to the cluster. What one maybe needs to do is to separate `deploy` and `retrieve` better in the sense that `deploy` only deploys the cluster and maybe gives a handle to shut down the cluster using a native client (e.g. `YarnClient` or `KubeClient`) and then `retrieve` which tries to create the `RestClusterClient` for a cluster.
   
   Don't get me wrong here. I think your solution works as a quick fix and maybe that is what we should do. But I think that this problem shows that there is more of a conceptual problem with the overall design because we need special case logic for `ClusterIP`.


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org