I'm using apache spark 2.2.1, that running on Amazon EMR cluster. Sometimes jobs fail on 'Futures timed out':
java.util.concurrent.TimeoutException: Futures timed out after [100000 milliseconds]
at scala.concurrent.impl.Promise$DefaultPromise.ready(Promise.scala:219)
at scala.concurrent.impl.Promise$DefaultPromise.result(Promise.scala:223)
at org.apache.spark.util.ThreadUtils$.awaitResult(ThreadUtils.scala:201)
at org.apache.spark.deploy.yarn.ApplicationMaster.runDriver(ApplicationMaster.scala:401)
at org.apache.spark.deploy.yarn.ApplicationMaster.run(ApplicationMaster.scala:254)
at org.apache.spark.deploy.yarn.ApplicationMaster$$anonfun$main$1.apply$mcV$sp(ApplicationMaster.scala:764)
at org.apache.spark.deploy.SparkHadoopUtil$$anon$2.run(SparkHadoopUtil.scala:67)
at org.apache.spark.deploy.SparkHadoopUtil$$anon$2.run(SparkHadoopUtil.scala:66)
at java.security.AccessController.doPrivileged(Native Method)
at javax.security.auth.Subject.doAs(Subject.java:422)
at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1836)
at org.apache.spark.deploy.SparkHadoopUtil.runAsSparkUser(SparkHadoopUtil.scala:66)
at org.apache.spark.deploy.yarn.ApplicationMaster$.main(ApplicationMaster.scala:762)
at org.apache.spark.deploy.yarn.ApplicationMaster.main(ApplicationMaster.scala)
I changed 2 params in spark-defaults.conf:
spark.sql.broadcastTimeout 1000
spark.network.timeout 10000000
but it didn't help.
Do you have any suggestions on how to handle this timeout?
Have you tried setting spark.yarn.am.waitTime?
Only used in cluster mode. Time for the YARN Application Master to
wait for the SparkContext to be initialized.
The quote above is from here.
A bit more context on my situation:
I am using spark-submit to execute a java-spark job. I deploy the client to the cluster, and the client is doing a very long running operation which was causing a time out.
I got around it by:
spark-submit --master yarn --deploy-mode cluster --conf "spark.yarn.am.waitTime=600000"
Related
My spark submit syntax is:
spark-submit --queue regular --deploy-mode cluster --conf spark.locality.wait=5000000ms --num-executors 100 --executor-memory 40G job.py
And the exception happened after job run successfully for a while, which is:
Application diagnostics message: User application exited with status 1
Exception in thread "main" org.apache.spark.SparkException: Application application_1635856758535_5228470 finished with failed status
at org.apache.spark.deploy.yarn.Client.run(Client.scala:1150)
at org.apache.spark.deploy.yarn.YarnClusterApplication.start(Client.scala:1530)
at org.apache.spark.deploy.SparkSubmit.org$apache$spark$deploy$SparkSubmit$$runMain(SparkSubmit.scala:845)
at org.apache.spark.deploy.SparkSubmit.doRunMain$1(SparkSubmit.scala:161)
at org.apache.spark.deploy.SparkSubmit.submit(SparkSubmit.scala:184)
at org.apache.spark.deploy.SparkSubmit.doSubmit(SparkSubmit.scala:86)
at org.apache.spark.deploy.SparkSubmit$$anon$2.doSubmit(SparkSubmit.scala:920)
at org.apache.spark.deploy.SparkSubmit$.main(SparkSubmit.scala:929)
at org.apache.spark.deploy.SparkSubmit.main(SparkSubmit.scala)
More importantly, in the job.py script, there's no local file read or write. Only parquet read or write.
And I cannot even track the application UI as the application was closed because of the exception. If anyone else encountered similar issue and has any ideas, please advice any solutions. Thanks!
I have spark streaming application using checkpoint writing on HDFS.
Has anyone know the solution?
Previously we were using the kinit to specify principal and keytab and got the suggestion to specify these via spark-submit command instead kinit but still this error and cause spark streaming application down.
spark-submit --principal sparkuser#HADOOP.ABC.COM --keytab /home/sparkuser/keytab/sparkuser.keytab --name MyStreamingApp --master yarn-cluster --conf "spark.driver.extraJavaOptions=-XX:+UseConcMarkSweepGC --conf "spark.eventLog.enabled=true" --conf "spark.streaming.backpressure.enabled=true" --conf "spark.streaming.stopGracefullyOnShutdown=true" --conf "spark.executor.extraJavaOptions=-XX:+UseConcMarkSweepGC --class com.abc.DataProcessor myapp.jar
I see multiple occurrences of following exception in logs and finally SIGTERM 15 that kills the executor and driver. We are using CDH 5.5.2
2016-10-02 23:59:50 ERROR SparkListenerBus LiveListenerBus:96 -
Listener EventLoggingListener threw an exception
java.lang.reflect.InvocationTargetException
at sun.reflect.GeneratedMethodAccessor8.invoke(Unknown Source)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
at java.lang.reflect.Method.invoke(Method.java:606)
at org.apache.spark.scheduler.EventLoggingListener$$anonfun$logEvent$3.apply(EventLoggingListener.scala:148)
at org.apache.spark.scheduler.EventLoggingListener$$anonfun$logEvent$3.apply(EventLoggingListener.scala:148)
at scala.Option.foreach(Option.scala:236)
at org.apache.spark.scheduler.EventLoggingListener.logEvent(EventLoggingListener.scala:148)
at org.apache.spark.scheduler.EventLoggingListener.onUnpersistRDD(EventLoggingListener.scala:184)
at org.apache.spark.scheduler.SparkListenerBus$class.onPostEvent(SparkListenerBus.scala:50)
at org.apache.spark.scheduler.LiveListenerBus.onPostEvent(LiveListenerBus.scala:31)
at org.apache.spark.scheduler.LiveListenerBus.onPostEvent(LiveListenerBus.scala:31)
at org.apache.spark.util.ListenerBus$class.postToAll(ListenerBus.scala:56)
at org.apache.spark.util.AsynchronousListenerBus.postToAll(AsynchronousListenerBus.scala:37)
at org.apache.spark.util.AsynchronousListenerBus$$anon$1$$anonfun$run$1.apply$mcV$sp(AsynchronousListenerBus.scala:79)
at org.apache.spark.util.Utils$.tryOrStopSparkContext(Utils.scala:1135)
at org.apache.spark.util.AsynchronousListenerBus$$anon$1.run(AsynchronousListenerBus.scala:63)
Caused by: java.io.IOException: Lease timeout of 0 seconds expired.
at org.apache.hadoop.hdfs.DFSOutputStream.abort(DFSOutputStream.java:2370)
at org.apache.hadoop.hdfs.DFSClient.closeAllFilesBeingWritten(DFSClient.java:964)
at org.apache.hadoop.hdfs.DFSClient.renewLease(DFSClient.java:932)
at org.apache.hadoop.hdfs.LeaseRenewer.renew(LeaseRenewer.java:423)
at org.apache.hadoop.hdfs.LeaseRenewer.run(LeaseRenewer.java:448)
at org.apache.hadoop.hdfs.LeaseRenewer.access$700(LeaseRenewer.java:71)
at org.apache.hadoop.hdfs.LeaseRenewer$1.run(LeaseRenewer.java:304)
at java.lang.Thread.run(Thread.java:745)
I'm running Spark in a standalone cluster where spark master, worker and submit each run in there own Docker container.
When spark-submit my Java App with the --repositories and --packages options I can see that it successfully downloads the apps required dependencies. However the stderr logs on the spark workers web ui reports a java.lang.ClassNotFoundException: kafka.serializer.StringDecoder. This class is available in one of the dependencies downloaded by spark-submit. But doesn't look like it's available on the worker classpath??
16/02/22 16:17:09 INFO SparkDeploySchedulerBackend: SchedulerBackend is ready for scheduling beginning after reached minRegisteredResourcesRatio: 0.0
Exception in thread "main" java.lang.reflect.InvocationTargetException
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
at java.lang.reflect.Method.invoke(Method.java:498)
at org.apache.spark.deploy.worker.DriverWrapper$.main(DriverWrapper.scala:58)
at org.apache.spark.deploy.worker.DriverWrapper.main(DriverWrapper.scala)
Caused by: java.lang.NoClassDefFoundError: kafka/serializer/StringDecoder
at com.my.spark.app.JavaDirectKafkaWordCount.main(JavaDirectKafkaWordCount.java:71)
... 6 more
Caused by: java.lang.ClassNotFoundException: kafka.serializer.StringDecoder
at java.net.URLClassLoader.findClass(URLClassLoader.java:381)
at java.lang.ClassLoader.loadClass(ClassLoader.java:424)
at java.lang.ClassLoader.loadClass(ClassLoader.java:357)
... 7 more
The spark-submit call:
${SPARK_HOME}/bin/spark-submit --deploy-mode cluster \
--master spark://spark-master:7077 \
--repositories https://oss.sonatype.org/content/groups/public/ \
--packages org.apache.spark:spark-streaming-kafka_2.10:1.6.0,org.elasticsearch:elasticsearch-spark_2.10:2.2.0 \
--class com.my.spark.app.JavaDirectKafkaWordCount \
/app/spark-app.jar kafka-server:9092 mytopic
I was working with Spark 2.4.0 when I ran into this problem. I don't have a solution yet but just some observations based on experimentation and reading around for solutions. I am noting them down here just in case it helps some one in their investigation. I will update this answer if I find more information later.
The --repositories option is required only if some custom repository has to be referenced
By default the maven central repository is used if the --repositories option is not provided
When --packages option is specified, the submit operation tries to look for the packages and their dependencies in the ~/.ivy2/cache, ~/.ivy2/jars, ~/.m2/repository directories.
If they are not found, then they are downloaded from maven central using ivy and stored under the ~/.ivy2 directory.
In my case I had observed that
spark-shell worked perfectly with the --packages option
spark-submit would fail to do the same. It would download the dependencies correctly but fail to pass on the jars to the driver and worker nodes
spark-submit worked with the --packages option if I ran the driver locally using --deploy-mode client instead of cluster.
This would run the driver locally in the command shell where I ran the spark-submit command but the worker would run on the cluster with the appropriate dependency jars
I found the following discussion useful but I still have to nail down this problem.
https://github.com/databricks/spark-redshift/issues/244#issuecomment-347082455
Most people just use an UBER jar to avoid running into this problem and even to avoid the problem of conflicting jar versions where a different version of the same dependency jar is provided by the platform.
But I don't like that idea beyond a stop gap arrangement and am still looking for a solution.
I am using Horton Works Cluster (2 Node cluster) to run the spark and flume , So when I am running the job with --master "local[*]" , Flume is able to send the events and Spark is also able to receive and on checking at localhost:4040 I can see the events are being received from the flume. (We are pumping 100 Events/Sec from flume using flume-ng-sql source with an approx size of ~1KB each)
Where as when I run the same example with --master "yarn-client" , I am getting the below error in flume and spark is not getting any events as well.
2015-08-13 18:24:24,927 (SinkRunner-PollingRunner-DefaultSinkProcessor) [ERROR - org.apache.flume.SinkRunner$PollingRunner.run(SinkRunner.java:160)] Unable to deliver event. Exception follows.
org.apache.flume.EventDeliveryException: Failed to send events
at org.apache.flume.sink.AbstractRpcSink.process(AbstractRpcSink.java:403)
at org.apache.flume.sink.DefaultSinkProcessor.process(DefaultSinkProcessor.java:68)
at org.apache.flume.SinkRunner$PollingRunner.run(SinkRunner.java:147)
at java.lang.Thread.run(Thread.java:745)
Caused by: org.apache.flume.FlumeException: NettyAvroRpcClient { host: localhost, port: 55555 }: RPC connection error
at org.apache.flume.api.NettyAvroRpcClient.connect(NettyAvroRpcClient.java:182)
at org.apache.flume.api.NettyAvroRpcClient.connect(NettyAvroRpcClient.java:121)
at org.apache.flume.api.NettyAvroRpcClient.configure(NettyAvroRpcClient.java:638)
at org.apache.flume.api.RpcClientFactory.getInstance(RpcClientFactory.java:88)
at org.apache.flume.sink.AvroSink.initializeRpcClient(AvroSink.java:127)
at org.apache.flume.sink.AbstractRpcSink.createConnection(AbstractRpcSink.java:222)
at org.apache.flume.sink.AbstractRpcSink.verifyConnection(AbstractRpcSink.java:283)
at org.apache.flume.sink.AbstractRpcSink.process(AbstractRpcSink.java:360)
... 3 more
Caused by: java.io.IOException: Error connecting to localhost/127.0.0.1:55555
at org.apache.avro.ipc.NettyTransceiver.getChannel(NettyTransceiver.java:261)
at org.apache.avro.ipc.NettyTransceiver.<init>(NettyTransceiver.java:203)
at org.apache.avro.ipc.NettyTransceiver.<init>(NettyTransceiver.java:152)
at org.apache.flume.api.NettyAvroRpcClient.connect(NettyAvroRpcClient.java:168)
... 10 more
Caused by: java.net.ConnectException: Connection refused
at sun.nio.ch.SocketChannelImpl.checkConnect(Native Method)
at sun.nio.ch.SocketChannelImpl.finishConnect(SocketChannelImpl.java:744)
at org.jboss.netty.channel.socket.nio.NioClientSocketPipelineSink$Boss.connect(NioClientSocketPipelineSink.java:496)
at org.jboss.netty.channel.socket.nio.NioClientSocketPipelineSink$Boss.processSelectedKeys(NioClientSocketPipelineSink.java:452)
at org.jboss.netty.channel.socket.nio.NioClientSocketPipelineSink$Boss.run(NioClientSocketPipelineSink.java:365)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
... 1 more
^
Also below observation has been observed in cluster:
-- Memory consumption using yarn is pretty much higher than compared to that being used in case of local.
-- Also when I am pumping 100 events per 30 second then Flume and spark are able to connect and process the same using yarn-client as well as local..
Below is the command which I am using for flume and spark.
Flume:
sudo -u hdfs flume-ng agent --conf conf/ -f conf/flume_mysql_spark.conf -n agent1 -Dflume.root.logger=INFO,console > flumelog.txt
Spark:
sudo -u hdfs spark-submit --master "yarn-client" --class "org.paladion.atm.FlumeEventCount" target/atm-1.1-jar-with-dependencies.jar > sparklog.txt
sudo -u hdfs spark-submit --master "local[*]" --class "org.paladion.atm.FlumeEventCount" target/atm-1.1-jar-with-dependencies.jar > sparklog.txt
Kindly l;et me know what could be wrong over here?
It got solves as below:
1 - If running as local give IP of local machine in Flume as well as spark.
2 - If running as cluster (yarn-client or yarn-cluster) give IP of the machine in cluster where you want to send the events (other than the one where you are executing the program so may be give IP of node which is not a master node) machine in Flume as well as spark.
Let me know if I am wrong and this could have worked for some other reason and any better solution is there for the same.
On running this command:
~/spark/bin/spark-submit --class [class-name] --master [spark-master-url]:7077 [jar-path]
I am getting
java.lang.RuntimeException: java.net.ConnectException: Call to ec2-[ip].compute-1.amazonaws.com/[internal-ip]:9000 failed on connection exception: java.net.ConnectException: Connection refused
Using spark version 1.3.0.
How do I resolve it?
When Spark is run in Cluster mode, all input files will be expected to be from HDFS (otherwise how will workers read from master's local files). But in this case, Hadoop wasn't running, so it was giving this exception.
Starting HDFS resolved this.