When flink performs a checkpoint I get randomly ConcurrentModificationException.
>From my investigation it looks like the method public boolean advance() throws IOException <goog_859136059> from https://github.com/apache/beam/blob/release-2.0.0/sdks/java/io/kinesis/src/main/java/org/apache/beam/sdk/io/kinesis/KinesisReader.java is called in another thread while checkpoint is being performed. The exception is caused because the method public UnboundedSource.CheckpointMark getCheckpointMark() from the KinesisReader.java <https://github.com/apache/beam/blob/release-2.0.0/sdks/java/io/kinesis/src/main/java/org/apache/beam/sdk/io/kinesis/KinesisReader.java> is iterating over iterator returned by RoundRobin<ShardRecordsIterator>.iterator() while a public boolean advance() throws IOException is calling RoundRobin<ShardRecordsIterator>.moveForward() from another thread which is causing java.util.ConcurrentModificationException to be thrown. RoundRobin<ShardRecordsIterator> class is using java.util.Deque queue which doesn't allow adding/removal of element while it's being iterated. Is some locking missing? I am using Beam 2.0.0, Flink 1.2.1, 20 slots and 32 kinesis shards. I created a bug for it as well https://issues.apache.org/jira/browse/BEAM-2752 Stacktrace: java.lang.Exception: Error while triggering checkpoint 59 for Source: Read(KinesisSource) -> Flat Map -> ParMultiDo(KinesisExtractor) -> Flat Map -> ParMultiDo(StringToRecord) -> Flat Map -> ParMultiDo(Anonymous) -> Flat Map -> ParMultiDo(ToRRecord) -> Flat Map -> ParMultiDo(AddTimestamps) -> Flat Map -> xxxx.yyyy.GroupByOneMinuteWindow GROUP RDOTRECORDS BY ONE MINUTE WINDOWS/Window.Assign.out -> (ParMultiDo(Anonymous) -> Flat Map -> ParMultiDo(ToSomeKey) -> Flat Map -> ToKeyedWorkItem, ParMultiDo(ToCompositeKey) -> Flat Map -> ParMultiDo(Anonymous) -> Flat Map -> ToKeyedWorkItem, ParMultiDo(Anonymous) -> Flat Map -> ParMultiDo(ApplyShardingKey) -> Flat Map -> ToKeyedWorkItem) (1/20) at org.apache.flink.runtime.taskmanager.Task$3.run(Task.java:1136) at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511) at java.util.concurrent.FutureTask.run(FutureTask.java:266) at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149) at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624) at java.lang.Thread.run(Thread.java:748) Caused by: java.lang.Exception: Could not perform checkpoint 59 for operator Source: Read(KinesisSource) -> Flat Map -> ParMultiDo(KinesisExtractor) -> Flat Map -> ParMultiDo(StringToRecord) -> Flat Map -> ParMultiDo(Anonymous) -> Flat Map -> ParMultiDo(ToRRecord) -> Flat Map -> ParMultiDo(AddTimestamps) -> Flat Map -> xxxx.yyyy.GroupByOneMinuteWindow GROUP RDOTRECORDS BY ONE MINUTE WINDOWS/Window.Assign.out -> (ParMultiDo(Anonymous) -> Flat Map -> ParMultiDo(ToSomeKey) -> Flat Map -> ToKeyedWorkItem, ParMultiDo(ToCompositeKey) -> Flat Map -> ParMultiDo(Anonymous) -> Flat Map -> ToKeyedWorkItem, ParMultiDo(Anonymous) -> Flat Map -> ParMultiDo(ApplyShardingKey) -> Flat Map -> ToKeyedWorkItem) (1/20). at org.apache.flink.streaming.runtime.tasks.StreamTask.triggerCheckpoint(StreamTask.java:524) at org.apache.flink.runtime.taskmanager.Task$3.run(Task.java:1125) ... 5 more Caused by: java.lang.Exception: Could not complete snapshot 59 for operator Source: Read(KinesisSource) -> Flat Map -> ParMultiDo(KinesisExtractor) -> Flat Map -> ParMultiDo(StringToRecord) -> Flat Map -> ParMultiDo(Anonymous) -> Flat Map -> ParMultiDo(ToRRecord) -> Flat Map -> ParMultiDo(AddTimestamps) -> Flat Map -> xxxx.yyyy.GroupByOneMinuteWindow GROUP RDOTRECORDS BY ONE MINUTE WINDOWS/Window.Assign.out -> (ParMultiDo(Anonymous) -> Flat Map -> ParMultiDo(ToSomeKey) -> Flat Map -> ToKeyedWorkItem, ParMultiDo(ToCompositeKey) -> Flat Map -> ParMultiDo(Anonymous) -> Flat Map -> ToKeyedWorkItem, ParMultiDo(Anonymous) -> Flat Map -> ParMultiDo(ApplyShardingKey) -> Flat Map -> ToKeyedWorkItem) (1/20). at org.apache.flink.streaming.api.operators.AbstractStreamOperator.snapshotState(AbstractStreamOperator.java:379) at org.apache.flink.streaming.runtime.tasks.StreamTask$CheckpointingOperation.checkpointStreamOperator(StreamTask.java:1157) at org.apache.flink.streaming.runtime.tasks.StreamTask$CheckpointingOperation.executeCheckpointing(StreamTask.java:1090) at org.apache.flink.streaming.runtime.tasks.StreamTask.checkpointState(StreamTask.java:630) at org.apache.flink.streaming.runtime.tasks.StreamTask.performCheckpoint(StreamTask.java:575) at org.apache.flink.streaming.runtime.tasks.StreamTask.triggerCheckpoint(StreamTask.java:518) ... 6 more Caused by: java.util.ConcurrentModificationException at java.util.ArrayDeque$DeqIterator.next(ArrayDeque.java:643) at org.apache.beam.sdks.java.io.kinesis.repackaged.com.google.common.collect.TransformedIterator.next(TransformedIterator.java:47) at org.apache.beam.sdks.java.io.kinesis.repackaged.com.google.common.collect.ImmutableCollection$Builder.addAll(ImmutableCollection.java:409) at org.apache.beam.sdks.java.io.kinesis.repackaged.com.google.common.collect.ImmutableList$Builder.addAll(ImmutableList.java:699) at org.apache.beam.sdks.java.io.kinesis.repackaged.com.google.common.collect.ImmutableList.copyOf(ImmutableList.java:256) at org.apache.beam.sdks.java.io.kinesis.repackaged.com.google.common.collect.ImmutableList.copyOf(ImmutableList.java:209) at org.apache.beam.sdk.io.kinesis.KinesisReaderCheckpoint.<init>(KinesisReaderCheckpoint.java:44) at org.apache.beam.sdk.io.kinesis.KinesisReaderCheckpoint.asCurrentStateOf(KinesisReaderCheckpoint.java:49) at org.apache.beam.sdk.io.kinesis.KinesisReader.getCheckpointMark(KinesisReader.java:137) at org.apache.beam.runners.flink.translation.wrappers.streaming.io.UnboundedSourceWrapper.snapshotState(UnboundedSourceWrapper.java:379) at org.apache.flink.streaming.api.functions.util.StreamingFunctionUtils.trySnapshotFunctionState(StreamingFunctionUtils.java:118) at org.apache.flink.streaming.api.functions.util.StreamingFunctionUtils.snapshotFunctionState(StreamingFunctionUtils.java:99) at org.apache.flink.streaming.api.operators.AbstractUdfStreamOperator.snapshotState(AbstractUdfStreamOperator.java:100) at org.apache.flink.streaming.api.operators.AbstractStreamOperator.snapshotState(AbstractStreamOperator.java:357) ... 11 more
