Github user EAlexRojas commented on a diff in the pull request: https://github.com/apache/flink/pull/5991#discussion_r190519624 --- Diff: flink-connectors/flink-connector-kafka-base/src/main/java/org/apache/flink/streaming/connectors/kafka/FlinkKafkaConsumerBase.java --- @@ -235,7 +243,8 @@ public FlinkKafkaConsumerBase( Pattern topicPattern, KeyedDeserializationSchema<T> deserializer, long discoveryIntervalMillis, - boolean useMetrics) { + boolean useMetrics, + boolean checkUnavailablePartitions) { --- End diff -- I did it in that way only because this is something new, so I though that maybe you may want it to be configurable. But you are right I cannot think of a case we would prefer to keep the unavailable partitions. I'll update the PR to make it the default behaviour if it's ok for you.
---