Github user tzulitai commented on a diff in the pull request:
https://github.com/apache/flink/pull/5991#discussion_r190153528
--- Diff:
flink-connectors/flink-connector-kafka-0.9/src/main/java/org/apache/flink/streaming/connectors/kafka/internal/KafkaConsumerThread.java
---
@@ -80,6 +83,9 @@
/** The queue of unassigned partitions that we need to assign to the
Kafka consumer. */
private final
ClosableBlockingQueue<KafkaTopicPartitionState<TopicPartition>>
unassignedPartitionsQueue;
+ /** The list of partitions to be removed from kafka consumer. */
+ private final Set<TopicPartition> partitionsToBeRemoved;
--- End diff --
Would it actually make more sense that we have a queue for this? Like how
we are handling unassigned new partitions via the `unassignedPartitionsQueue`.
The fact that this is a set means that we will need to eventually remove
entries from it anyways.
---
