Github user tzulitai commented on a diff in the pull request:
https://github.com/apache/flink/pull/3001#discussion_r112802258
--- Diff:
flink-connectors/flink-connector-kinesis/src/main/java/org/apache/flink/streaming/connectors/kinesis/FlinkKinesisConsumer.java
---
@@ -267,38 +289,77 @@ public void close() throws Exception {
//
------------------------------------------------------------------------
@Override
- public HashMap<KinesisStreamShard, SequenceNumber> snapshotState(long
checkpointId, long checkpointTimestamp) throws Exception {
+ public void snapshotState(FunctionSnapshotContext context) throws
Exception {
if (lastStateSnapshot == null) {
LOG.debug("snapshotState() requested on not yet opened
source; returning null.");
- return null;
- }
-
- if (fetcher == null) {
+ } else if (fetcher == null) {
LOG.debug("snapshotState() requested on not yet running
source; returning null.");
- return null;
- }
-
- if (!running) {
+ } else if (!running) {
LOG.debug("snapshotState() called on closed source;
returning null.");
- return null;
- }
+ } else {
+ if (LOG.isDebugEnabled()) {
+ LOG.debug("Snapshotting state ...");
+ }
- if (LOG.isDebugEnabled()) {
- LOG.debug("Snapshotting state ...");
- }
+ sequenceNumsStateForCheckpoint.clear();
+ lastStateSnapshot = fetcher.snapshotState();
- lastStateSnapshot = fetcher.snapshotState();
+ if (LOG.isDebugEnabled()) {
+ LOG.debug("Snapshotted state, last processed
sequence numbers: {}, checkpoint id: {}, timestamp: {}",
+ lastStateSnapshot.toString(),
context.getCheckpointId(), context.getCheckpointTimestamp());
+ }
- if (LOG.isDebugEnabled()) {
- LOG.debug("Snapshotted state, last processed sequence
numbers: {}, checkpoint id: {}, timestamp: {}",
- lastStateSnapshot.toString(), checkpointId,
checkpointTimestamp);
+ for (Map.Entry<KinesisStreamShard, SequenceNumber>
entry : lastStateSnapshot.entrySet()) {
+
sequenceNumsStateForCheckpoint.add(Tuple2.of(entry.getKey(), entry.getValue()));
+ }
}
+ }
- return lastStateSnapshot;
+ @Override
+ public void initializeState(FunctionInitializationContext context)
throws Exception {
+ TypeInformation<Tuple2<KinesisStreamShard, SequenceNumber>>
tuple = new TupleTypeInfo<>(
+ TypeInformation.of(KinesisStreamShard.class),
+ TypeInformation.of(SequenceNumber.class)
+ );
+
+ sequenceNumsStateForCheckpoint =
context.getOperatorStateStore().getUnionListState(
+ new ListStateDescriptor<>("Kinesis-Stream-Shard-State",
tuple));
+
+ if (context.isRestored()) {
+ if (sequenceNumsToRestore == null) {
+ sequenceNumsToRestore = new HashMap<>();
+ for (Tuple2<KinesisStreamShard, SequenceNumber>
kinesisOffset : sequenceNumsStateForCheckpoint.get()) {
+
sequenceNumsToRestore.put(kinesisOffset.f0, kinesisOffset.f1);
+ }
+
+ LOG.info("Setting restore state in the
FlinkKinesisConsumer. Using the following offsets: {}",
+ sequenceNumsToRestore);
+ } else if (sequenceNumsToRestore.isEmpty()) {
+ sequenceNumsToRestore = null;
+ }
+ } else {
+ LOG.info("No restore state for FlinkKinesisConsumer.");
+ }
}
@Override
public void restoreState(HashMap<KinesisStreamShard, SequenceNumber>
restoredState) throws Exception {
- sequenceNumsToRestore = restoredState;
+ LOG.info("Subtask {} restoring offsets from an older Flink
version: {}",
+ getRuntimeContext().getIndexOfThisSubtask(),
sequenceNumsToRestore);
+
+ sequenceNumsToRestore = restoredState.isEmpty() ? null :
restoredState;
+ }
+
+ protected KinesisDataFetcher<T> createFetcher(List<String> streams,
--- End diff --
Maybe a comment that this is exposed for test mocks would be nice.
---
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 [email protected] or file a JIRA ticket
with INFRA.
---
