Github user tzulitai commented on a diff in the pull request:
https://github.com/apache/flink/pull/2509#discussion_r89755327
--- Diff:
flink-streaming-connectors/flink-connector-kafka-0.10/src/main/java/org/apache/flink/streaming/connectors/kafka/FlinkKafkaConsumer010.java
---
@@ -128,6 +129,7 @@ public FlinkKafkaConsumer010(List<String> topics,
KeyedDeserializationSchema<T>
protected AbstractFetcher<T, ?> createFetcher(
SourceContext<T> sourceContext,
List<KafkaTopicPartition> thisSubtaskPartitions,
+ HashMap<KafkaTopicPartition, Long>
restoredSnapshotState,
--- End diff --
To make the startup mode logic cleaner, I've changed the `AbstractFetcher`
life cycle a bit.
Now, restored state is provided when constructing the `AbstractFetcher`,
instead of explicitly calling `AbstractFetcher#restoreOffsets()` as a separate
call.
This allows the AbstractFetcher to have a final `isRestored` flag that
version-specific implementations can use.
The startup offset configuring logic is much simpler now with this flag:
```
if (isRestored) {
// all subscribed partition states should have defined offset
// setup the KafkaConsumer client we're using to respect these restored
offsets
} else {
// all subscribed partition states have no defined offset
// (1) set offsets depending on whether startup mode is EARLIEST, LATEST,
or GROUP_OFFSET
// (2) use the fetched offsets from Kafka to set the initial partition
states we use in Flink.
}
```
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
