tillrohrmann commented on a change in pull request #14692:
URL: https://github.com/apache/flink/pull/14692#discussion_r560130091
##########
File path:
flink-kubernetes/src/main/java/org/apache/flink/kubernetes/KubernetesClusterDescriptor.java
##########
@@ -125,6 +121,16 @@ public String getClusterDescription() {
};
}
+ private String getWebMonitorAddress(Configuration configuration) throws
Exception {
+ HighAvailabilityServicesUtils.AddressResolution resolution =
+
HighAvailabilityServicesUtils.AddressResolution.TRY_ADDRESS_RESOLUTION;
+ if
(configuration.get(KubernetesConfigOptions.REST_SERVICE_EXPOSED_TYPE)
+ == KubernetesConfigOptions.ServiceExposedType.ClusterIP) {
+ resolution =
HighAvailabilityServicesUtils.AddressResolution.NO_ADDRESS_RESOLUTION;
+ }
+ return
HighAvailabilityServicesUtils.getWebMonitorAddress(configuration, resolution);
+ }
Review comment:
Assuming the submission happens outside of the K8s cluster with
`ClusterIP` configured, does this mean that the `RestClusterClient` cannot talk
to the cluster? If yes, then there is not really a need for creating it, does
it? But then there is the problem that we could also deploy the cluster from
within the K8s cluster where the `RestClusterClient` can talk to the Flink
cluster, right?
Should we maybe say that user's have to submit the cluster in detached mode
if they cannot connect to it? The problem with the current solution is that we
might start an attached per job cluster with which we cannot talk. Hence, we
cannot request the final job result either and consequently, the cluster will
never shut down. So maybe we should fail with an better exception message, what
do you think?
----------------------------------------------------------------
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:
us...@infra.apache.org
