使用Flink时遇到的问题（不断更新中）

1.启动不起来

查看JobManager日志：

WARN  org.apache.flink.runtime.webmonitor.JobManagerRetriever       - Failed to retrieve leader gateway and port.
akka.actor.ActorNotFound: Actor not found for: ActorSelection[Anchor(akka.tcp://flink@t-sha1-flk-01:6123/), Path(/user/jobmanager)]
    at akka.actor.ActorSelection$$anonfun$resolveOne$.apply(ActorSelection.scala:)
    at akka.actor.ActorSelection$$anonfun$resolveOne$.apply(ActorSelection.scala:)
    at scala.concurrent.impl.CallbackRunnable.run(Promise.scala:)
    at akka.dispatch.BatchingExecutor$AbstractBatch.processBatch(BatchingExecutor.scala:)
    at akka.dispatch.BatchingExecutor$Batch.run(BatchingExecutor.scala:)
    at akka.dispatch.ExecutionContexts$sameThreadExecutionContext$.unbatchedExecute(Future.scala:)
    at akka.dispatch.BatchingExecutor$class.execute(BatchingExecutor.scala:)
    at akka.dispatch.ExecutionContexts$sameThreadExecutionContext$.execute(Future.scala:)
    at scala.concurrent.impl.CallbackRunnable.executeWithValue(Promise.scala:)
    at scala.concurrent.impl.Promise$DefaultPromise.tryComplete(Promise.scala:)
    at akka.pattern.PromiseActorRef$$anonfun$.apply$mcV$sp(AskSupport.scala:)
    at akka.actor.Scheduler$$anon$.run(Scheduler.scala:)
    at scala.concurrent.Future$InternalCallbackExecutor$.unbatchedExecute(Future.scala:)
    at scala.concurrent.BatchingExecutor$class.execute(BatchingExecutor.scala:)
    at scala.concurrent.Future$InternalCallbackExecutor$.execute(Future.scala:)
    at akka.actor.LightArrayRevolverScheduler$TaskHolder.executeTask(Scheduler.scala:)
    at akka.actor.LightArrayRevolverScheduler$$anon$.executeBucket$(Scheduler.scala:)
    at akka.actor.LightArrayRevolverScheduler$$anon$.nextTick(Scheduler.scala:)
    at akka.actor.LightArrayRevolverScheduler$$anon$.run(Scheduler.scala:)
    at java.lang.Thread.run(Thread.java:)

解决方案：/etc/hosts中配置的主机名都是小写，但是在Flink配置文件（flink-config.yaml、masters、slaves）中配置的都是大写的hostname，将flink配置文件中的hostname都改为小写或者IP地址

2.运行一段时间退出

AsynchronousException{java.lang.Exception: Could not materialize checkpoint  for operator Compute By Event Time (/).}
    at org.apache.flink.streaming.runtime.tasks.StreamTask$AsyncCheckpointRunnable.run(StreamTask.java:)
    at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:)
    at java.util.concurrent.FutureTask.run(FutureTask.java:)
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:)
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:)
    at java.lang.Thread.run(Thread.java:)
Caused by: java.lang.Exception: Could not materialize checkpoint  for operator Compute By Event Time (/).
    ...  more
Caused by: java.util.concurrent.ExecutionException: java.io.IOException: Size of the state is larger than the maximum permitted memory-backed state. Size= , maxSize= . Consider using a different state backend, like the File System State backend.
    at java.util.concurrent.FutureTask.report(FutureTask.java:)
    at java.util.concurrent.FutureTask.get(FutureTask.java:)
    at org.apache.flink.util.FutureUtil.runIfNotDoneAndGet(FutureUtil.java:)
    at org.apache.flink.streaming.runtime.tasks.StreamTask$AsyncCheckpointRunnable.run(StreamTask.java:)
    ...  more
    Suppressed: java.lang.Exception: Could not properly cancel managed keyed state future.
        at org.apache.flink.streaming.api.operators.OperatorSnapshotResult.cancel(OperatorSnapshotResult.java:)
        at org.apache.flink.streaming.runtime.tasks.StreamTask$AsyncCheckpointRunnable.cleanup(StreamTask.java:)
        at org.apache.flink.streaming.runtime.tasks.StreamTask$AsyncCheckpointRunnable.run(StreamTask.java:)
        ...  more
    Caused by: java.util.concurrent.ExecutionException: java.io.IOException: Size of the state is larger than the maximum permitted memory-backed state. Size= , maxSize= . Consider using a different state backend, like the File System State backend.
        at java.util.concurrent.FutureTask.report(FutureTask.java:)
        at java.util.concurrent.FutureTask.get(FutureTask.java:)
        at org.apache.flink.util.FutureUtil.runIfNotDoneAndGet(FutureUtil.java:)
        at org.apache.flink.runtime.state.StateUtil.discardStateFuture(StateUtil.java:)
        at org.apache.flink.streaming.api.operators.OperatorSnapshotResult.cancel(OperatorSnapshotResult.java:)
        ...  more
    Caused by: java.io.IOException: Size of the state is larger than the maximum permitted memory-backed state. Size= , maxSize= . Consider using a different state backend, like the File System State backend.
        at org.apache.flink.runtime.state.memory.MemCheckpointStreamFactory.checkSize(MemCheckpointStreamFactory.java:)
        at org.apache.flink.runtime.state.memory.MemCheckpointStreamFactory$MemoryCheckpointOutputStream.closeAndGetBytes(MemCheckpointStreamFactory.java:)
        at org.apache.flink.runtime.state.memory.MemCheckpointStreamFactory$MemoryCheckpointOutputStream.closeAndGetHandle(MemCheckpointStreamFactory.java:)
        at org.apache.flink.runtime.checkpoint.AbstractAsyncSnapshotIOCallable.closeStreamAndGetStateHandle(AbstractAsyncSnapshotIOCallable.java:)
        at org.apache.flink.runtime.state.heap.HeapKeyedStateBackend$.performOperation(HeapKeyedStateBackend.java:)
        at org.apache.flink.runtime.state.heap.HeapKeyedStateBackend$.performOperation(HeapKeyedStateBackend.java:)
        at org.apache.flink.runtime.io.async.AbstractAsyncIOCallable.call(AbstractAsyncIOCallable.java:)
        at java.util.concurrent.FutureTask.run(FutureTask.java:)
        at org.apache.flink.runtime.state.heap.HeapKeyedStateBackend.snapshot(HeapKeyedStateBackend.java:)
        at org.apache.flink.streaming.api.operators.AbstractStreamOperator.snapshotState(AbstractStreamOperator.java:)
        at org.apache.flink.streaming.runtime.tasks.StreamTask$CheckpointingOperation.checkpointStreamOperator(StreamTask.java:)
        at org.apache.flink.streaming.runtime.tasks.StreamTask$CheckpointingOperation.executeCheckpointing(StreamTask.java:)
        at org.apache.flink.streaming.runtime.tasks.StreamTask.checkpointState(StreamTask.java:)
        at org.apache.flink.streaming.runtime.tasks.StreamTask.performCheckpoint(StreamTask.java:)
        at org.apache.flink.streaming.runtime.tasks.StreamTask.triggerCheckpointOnBarrier(StreamTask.java:)
        at org.apache.flink.streaming.runtime.io.BarrierBuffer.notifyCheckpoint(BarrierBuffer.java:)
        at org.apache.flink.streaming.runtime.io.BarrierBuffer.processBarrier(BarrierBuffer.java:)
        at org.apache.flink.streaming.runtime.io.BarrierBuffer.getNextNonBlocked(BarrierBuffer.java:)
        at org.apache.flink.streaming.runtime.io.StreamInputProcessor.processInput(StreamInputProcessor.java:)
        at org.apache.flink.streaming.runtime.tasks.OneInputStreamTask.run(OneInputStreamTask.java:)
        at org.apache.flink.streaming.runtime.tasks.StreamTask.invoke(StreamTask.java:)
        at org.apache.flink.runtime.taskmanager.Task.run(Task.java:)
        ...  more
    [CIRCULAR REFERENCE:java.io.IOException: Size of the state is larger than the maximum permitted memory-backed state. Size= , maxSize= . Consider using a different state backend, like the File System State backend.]

解决方案：

状态存储，默认是在内存中，改为存储到HDFS中：

state.backend.fs.checkpointdir: hdfs://t-sha1-flk-01:9000/flink-checkpoints

3.长时间运行后，多次重启

AsynchronousException{java.lang.Exception: Could not materialize checkpoint 1488 for operator Compute By Event Time -> (MonitorData, MonitorDataMapping, MonitorSamplingData) (6/6).}
at org.apache.flink.streaming.runtime.tasks.StreamTask$AsyncCheckpointRunnable.run(StreamTask.java:948)
at java.util.concurrent.Executors$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$Worker.run(ThreadPoolExecutor.java:624)
at java.lang.Thread.run(Thread.java:748)
Caused by: java.lang.Exception: Could not materialize checkpoint 1488 for operator Compute By Event Time -> (MonitorData, MonitorDataMapping, MonitorSamplingData) (6/6).
... 6 more
Caused by: java.util.concurrent.ExecutionException: org.apache.hadoop.ipc.RemoteException(java.io.IOException): File /flink-checkpoints/8c274785f1ab027e6146a59364be645f/chk-1488/2c612f30-c57d-4ede-9025-9554ca11fd12 could only be replicated to 0 nodes instead of minReplication (=1). There are 3 datanode(s) running and no node(s) are excluded in this operation.
at org.apache.hadoop.hdfs.server.blockmanagement.BlockManager.chooseTarget4NewBlock(BlockManager.java:1628)
at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.getNewBlockTargets(FSNamesystem.java:3121)
at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.getAdditionalBlock(FSNamesystem.java:3045)
at org.apache.hadoop.hdfs.server.namenode.NameNodeRpcServer.addBlock(NameNodeRpcServer.java:725)
at org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolServerSideTranslatorPB.addBlock(ClientNamenodeProtocolServerSideTranslatorPB.java:493)
at org.apache.hadoop.hdfs.protocol.proto.ClientNamenodeProtocolProtos$ClientNamenodeProtocol$2.callBlockingMethod(ClientNamenodeProtocolProtos.java)
at org.apache.hadoop.ipc.ProtobufRpcEngine$Server$ProtoBufRpcInvoker.call(ProtobufRpcEngine.java:616)
at org.apache.hadoop.ipc.RPC$Server.call(RPC.java:982)
at org.apache.hadoop.ipc.Server$Handler$1.run(Server.java:2217)
at org.apache.hadoop.ipc.Server$Handler$1.run(Server.java:2213)
at java.security.AccessController.doPrivileged(Native Method)
at javax.security.auth.Subject.doAs(Subject.java:422)
at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1746)
at org.apache.hadoop.ipc.Server$Handler.run(Server.java:2213)

查看hdfs日志，

WARN org.apache.hadoop.hdfs.protocol.BlockStoragePolicy: 
Failed to place enough replicas: expected size is 2 but only 0 storage types can be selected 
(replication=3, selected=[], unavailable=[DISK], removed=[DISK, DISK],
 policy=BlockStoragePolicy{HOT:7, storageTypes=[DISK], creationFallbacks=[], replicationFallbacks=[ARCHIVE]})

搭建的Flink使用HDFS作为CheckPoint的存储，当flink重启时，原来的checkpoint没有用了，我就手动给删了，不知道和这个有没有关系，为了不继续报异常，便重启了Flink、HDFS，重启后不再有异常信息了。

但是查看HDFS日志时，发现如下警告（不合规范的URI格式）：

WARN org.apache.hadoop.hdfs.server.common.Util:
Path /mnt/hadoop/dfs/name should be specified as a URI in configuration files. 
Please update hdfs configuration

原来是配置错了，hdfs-site.xml中的

    <property>
      <name>dfs.namenode.name.dir</name>
      <value>/mnt/hadoop/dfs/name</value>
    </property>

应该改为：

    <property>
      <name>dfs.namenode.name.dir</name>
      <value>file:/mnt/hadoop/dfs/name</value>
    </property>

至此问题解决，根上的问题应该是hdfs-site.xml配置的不对导致的。

4.Unable to load native-hadoop library for your platform

Flink启动时，有时会有如下警告信息：

WARN  org.apache.hadoop.util.NativeCodeLoader    

- Unable to load native-hadoop library for your platform... 

using builtin-java classes where applicable

参考资料1：http://blog.csdn.net/jack85986370/article/details/51902871

解决方案：编辑/etc/profile文件，增加

export HADOOP_COMMON_LIB_NATIVE_DIR=$HADOOP_HOME/lib/native

export HADOOP_OPTS="-Djava.library.path=$HADOOP_HOME/lib"

未能解决该问题

5.hadoop checknative -a

WARN bzip2.Bzip2Factory: Failed to load/initialize native-bzip2 library system-native, will use pure-Java version
INFO zlib.ZlibFactory: Successfully loaded & initialized native-zlib library
Native library checking:
hadoop:  true /usr/hadoop-2.7./lib/native/libhadoop.so.1.0.
zlib:    true /lib64/libz.so.
snappy:  false
lz4:     true revision:
bzip2:   false
openssl: false Cannot load libcrypto.so (libcrypto.so: cannot open shared object file: No such file or directory)!
INFO util.ExitUtil: Exiting with status 

参考资料：http://blog.csdn.net/zhangzhaokun/article/details/50951238

解决方案

 cd /usr/lib64/
 ln -s libcrypto.so.1.0.1e libcrypto.so

6.TaskManager退出

Flink运行一段时间后，出现TaskManager退出情况，通过jvisualvm抓取TaskManager的Dump，使用MAT进行分析，结果如下：

One instance of "org.apache.flink.runtime.io.network.buffer.NetworkBufferPool"
loaded by "sun.misc.Launcher$AppClassLoader @ 0x6c01de310" occupies 403,429,704 (76.24%) bytes. 
The memory is accumulated in one instance of "java.lang.Object[]" loaded by "<system class loader>".

Keywords
sun.misc.Launcher$AppClassLoader @ 0x6c01de310
java.lang.Object[]
org.apache.flink.runtime.io.network.buffer.NetworkBufferPool

发现是网络缓冲池不足，查到一篇文章：

https://issues.apache.org/jira/browse/FLINK-4536

和我遇到的情况差不多，也是使用了InfluxDB作为Sink，最后在Close里进行关闭，问题解决。

另外，在$FLINK_HOME/conf/flink-conf.yaml中，也有关于TaskManager网络栈的配置，暂时未调整

# The number of buffers for the network stack.
#
# taskmanager.network.numberOfBuffers: 

7.Kafka partition leader切换导致Flink重启

现象：

7.1 Flink重启，查看日志，显示：

java.lang.Exception: Failed to send data to Kafka: This server is not the leader for that topic-partition.
  at org.apache.flink.streaming.connectors.kafka.FlinkKafkaProducerBase.checkErroneous(FlinkKafkaProducerBase.java:)
  at org.apache.flink.streaming.connectors.kafka.FlinkKafkaProducerBase.invoke(FlinkKafkaProducerBase.java:)
  at org.apache.flink.streaming.api.operators.StreamSink.processElement(StreamSink.java:)
  at org.apache.flink.streaming.runtime.io.StreamInputProcessor.processInput(StreamInputProcessor.java:)
  at org.apache.flink.streaming.runtime.tasks.OneInputStreamTask.run(OneInputStreamTask.java:)
  at org.apache.flink.streaming.runtime.tasks.StreamTask.invoke(StreamTask.java:)
  at org.apache.flink.runtime.taskmanager.Task.run(Task.java:)
  at java.lang.Thread.run(Thread.java:)
Caused by: org.apache.kafka.common.errors.NotLeaderForPartitionException: This server is not the leader for that topic-partition.

7.2 查看Kafka的Controller日志，显示：

 INFO [SessionExpirationListener on ], ZK expired; shut down all controller components and 

try to re-elect (kafka.controller.KafkaController$SessionExpirationListener)

7.3 设置retries参数

参考：http://colabug.com/122248.html 以及 Kafka官方文档（http://kafka.apache.org/082/documentation.html#producerconfigs），关于producer参数设置

设置了retries参数，可以在Kafka的Partition发生leader切换时，Flink不重启，而是做3次尝试：

        kafkaProducerConfig
          {
                "bootstrap.servers": "192.169.2.20:9093,192.169.2.21:9093,192.169.2.22:9093"
                "retries":
          }