MDSNotificationFeedStore#85 logs an exception with the full stack trace when dataset service is unavailable. This keeps happening every second, so it is difficult to debug and find out why dataset service is not up. Maybe move the log message to debug? Or just have the message and move the exception to trace?
Also found the same at DistributedStreamService where it catches a NotificationException and logs the entire exception.
We could potentially use this pattern from DefaultMetricDatasetFactory to retry until dataset service is up too:
Ok I will use that same pattern.