Ran Tao created FLINK-8638: ------------------------------ Summary: Job restart when Checkpoint On Barrier failed Key: FLINK-8638 URL: https://issues.apache.org/jira/browse/FLINK-8638 Project: Flink Issue Type: Bug Components: Streaming Affects Versions: 1.3.2, 1.4.0 Reporter: Ran Tao
The following example comes from the one snapshotState process by using hdfs, snapshotState failed due to hdfs disk problems, so that triggerCheckpointOnBarrier fails and throws an exception to make the application restart. However, when restarting, flink needs to recover from the recent completed checkpoint and start chasing the data, which can lead to significant delays. We think that when StreamTask's triggerCheckpointOnBarrier (including the triggerCheckpoint at source) fails, the application should not restart but instead continue running and mark the checkpoint failed. Finally, notify the JobManager this checkpoint failed. By adding Checkpoint failure alarm let developers or users know this situation, and take the appropriate action. During this time, the flink job always keeps running. {code:java} java.lang.Exception: Could not perform checkpoint 45843 for operator TriggerWindow(TumblingEventTimeWindows(60000), ReducingStateDescriptor{serializer=com.didi.op.router.streaming.Launcher$$anon$49$$anon$13@af3ff050, reduceFunction=org.apache.flink.streaming.api.scala.function.util.ScalaReduceFunction@26f7cdf8}, EventTimeTrigger(), WindowedStream.reduce(WindowedStream.java:300)) -> Map (153/459). at org.apache.flink.streaming.runtime.tasks.StreamTask.triggerCheckpointOnBarrier(StreamTask.java:552) at org.apache.flink.streaming.runtime.io.BarrierBuffer.notifyCheckpoint(BarrierBuffer.java:378) at org.apache.flink.streaming.runtime.io.BarrierBuffer.processBarrier(BarrierBuffer.java:281) at org.apache.flink.streaming.runtime.io.BarrierBuffer.getNextNonBlocked(BarrierBuffer.java:183) at org.apache.flink.streaming.runtime.io.StreamInputProcessor.processInput(StreamInputProcessor.java:213) at org.apache.flink.streaming.runtime.tasks.OneInputStreamTask.run(OneInputStreamTask.java:69) at org.apache.flink.streaming.runtime.tasks.StreamTask.invoke(StreamTask.java:263) at org.apache.flink.runtime.taskmanager.Task.run(Task.java:702) at java.lang.Thread.run(Thread.java:745) Caused by: java.lang.Exception: Could not complete snapshot 45843 for operator TriggerWindow(TumblingEventTimeWindows(60000), ReducingStateDescriptor{serializer=com.didi.op.router.streaming.Launcher$$anon$49$$anon$13@af3ff050, reduceFunction=org.apache.flink.streaming.api.scala.function.util.ScalaReduceFunction@26f7cdf8}, EventTimeTrigger(), WindowedStream.reduce(WindowedStream.java:300)) -> Map (153/459). at org.apache.flink.streaming.api.operators.AbstractStreamOperator.snapshotState(AbstractStreamOperator.java:407) at org.apache.flink.streaming.runtime.tasks.StreamTask$CheckpointingOperation.checkpointStreamOperator(StreamTask.java:1162) at org.apache.flink.streaming.runtime.tasks.StreamTask$CheckpointingOperation.executeCheckpointing(StreamTask.java:1094) at org.apache.flink.streaming.runtime.tasks.StreamTask.checkpointState(StreamTask.java:654) at org.apache.flink.streaming.runtime.tasks.StreamTask.performCheckpoint(StreamTask.java:590) at org.apache.flink.streaming.runtime.tasks.StreamTask.triggerCheckpointOnBarrier(StreamTask.java:543) ... 8 more Caused by: java.io.IOException: Could not flush and close the file system output stream to hdfs://hadoop/checkpoint/flink/2406585590309979134/c8ac238c750d7b7b3e3b498bc396570f/chk-45843/5905b7a2-fc8c-4500-898a-7b87fa5470ee in order to obtain the stream state handle at org.apache.flink.runtime.state.filesystem.FsCheckpointStreamFactory$FsCheckpointStateOutputStream.closeAndGetHandle(FsCheckpointStreamFactory.java:336) at org.apache.flink.runtime.state.KeyedStateCheckpointOutputStream.closeAndGetHandle(KeyedStateCheckpointOutputStream.java:105) at org.apache.flink.runtime.state.KeyedStateCheckpointOutputStream.closeAndGetHandle(KeyedStateCheckpointOutputStream.java:30) at org.apache.flink.runtime.state.StateSnapshotContextSynchronousImpl.closeAndUnregisterStreamToObtainStateHandle(StateSnapshotContextSynchronousImpl.java:131) at org.apache.flink.runtime.state.StateSnapshotContextSynchronousImpl.getKeyedStateStreamFuture(StateSnapshotContextSynchronousImpl.java:113) at org.apache.flink.streaming.api.operators.AbstractStreamOperator.snapshotState(AbstractStreamOperator.java:387) ... 13 more Caused by: java.io.IOException: All datanodes DatanodeInfoWithStorage are bad. Aborting... at org.apache.hadoop.hdfs.DFSOutputStream$DataStreamer.setupPipelineForAppendOrRecovery(DFSOutputStream.java:1109) at org.apache.hadoop.hdfs.DFSOutputStream$DataStreamer.processDatanodeError(DFSOutputStream.java:871) at org.apache.hadoop.hdfs.DFSOutputStream$DataStreamer.run(DFSOutputStream.java:401){code} -- This message was sent by Atlassian JIRA (v7.6.3#76005)