You need to set a larger `spark.akka.frameSize`, e.g., 128, for the serialized weight vector. There is a JIRA about switching automatically between sending through akka or broadcast: https://issues.apache.org/jira/browse/SPARK-2361 . -Xiangrui
On Mon, Jul 14, 2014 at 12:15 AM, crater <cq...@ucmerced.edu> wrote: > Hi, > > I encounter an error when testing svm (example one) on very large sparse > data. The dataset I ran on was a toy dataset with only ten examples but 13 > million sparse vector with a few thousands non-zero entries. > > The errors is showing below. I am wondering is this a bug or I am missing > something? > > 14/07/13 23:59:44 INFO SecurityManager: Using Spark's default log4j profile: > org/apache/spark/log4j-defaults.properties > 14/07/13 23:59:44 INFO SecurityManager: Changing view acls to: chengjie > 14/07/13 23:59:44 INFO SecurityManager: SecurityManager: authentication > disabled; ui acls disabled; users with view permissions: Set(chengjie) > 14/07/13 23:59:45 INFO Slf4jLogger: Slf4jLogger started > 14/07/13 23:59:45 INFO Remoting: Starting remoting > 14/07/13 23:59:45 INFO Remoting: Remoting started; listening on addresses > :[akka.tcp://spark@master:53173] > 14/07/13 23:59:45 INFO Remoting: Remoting now listens on addresses: > [akka.tcp://spark@master:53173] > 14/07/13 23:59:45 INFO SparkEnv: Registering MapOutputTracker > 14/07/13 23:59:45 INFO SparkEnv: Registering BlockManagerMaster > 14/07/13 23:59:45 INFO DiskBlockManager: Created local directory at > /tmp/spark-local-20140713235945-c78f > 14/07/13 23:59:45 INFO MemoryStore: MemoryStore started with capacity 14.4 > GB. > 14/07/13 23:59:45 INFO ConnectionManager: Bound socket to port 37674 with id > = ConnectionManagerId(master,37674) > 14/07/13 23:59:45 INFO BlockManagerMaster: Trying to register BlockManager > 14/07/13 23:59:45 INFO BlockManagerInfo: Registering block manager > master:37674 with 14.4 GB RAM > 14/07/13 23:59:45 INFO BlockManagerMaster: Registered BlockManager > 14/07/13 23:59:45 INFO HttpServer: Starting HTTP Server > 14/07/13 23:59:45 INFO HttpBroadcast: Broadcast server started at > http://10.10.255.128:41838 > 14/07/13 23:59:45 INFO HttpFileServer: HTTP File server directory is > /tmp/spark-ac459d4b-a3c4-4577-bad4-576ac427d0bf > 14/07/13 23:59:45 INFO HttpServer: Starting HTTP Server > 14/07/13 23:59:51 INFO SparkUI: Started SparkUI at http://master:4040 > 14/07/13 23:59:51 WARN NativeCodeLoader: Unable to load native-hadoop > library for your platform... using builtin-java classes where applicable > 14/07/13 23:59:52 INFO EventLoggingListener: Logging events to > /tmp/spark-events/binaryclassification-with-params(hdfs---master-9001-splice.small,1,1.0,svm,l1,0.1)-1405317591776 > 14/07/13 23:59:52 INFO SparkContext: Added JAR > file:/home/chengjie/spark-1.0.1/examples/target/scala-2.10/spark-examples-1.0.1-hadoop2.3.0.jar > at http://10.10.255.128:54689/jars/spark-examples-1.0.1-hadoop2.3.0.jar with > timestamp 1405317592653 > 14/07/13 23:59:52 INFO AppClient$ClientActor: Connecting to master > spark://master:7077... > 14/07/14 00:00:08 WARN TaskSchedulerImpl: Initial job has not accepted any > resources; check your cluster UI to ensure that workers are registered and > have sufficient memory > 14/07/14 00:00:23 WARN TaskSchedulerImpl: Initial job has not accepted any > resources; check your cluster UI to ensure that workers are registered and > have sufficient memory > 14/07/14 00:00:38 WARN TaskSchedulerImpl: Initial job has not accepted any > resources; check your cluster UI to ensure that workers are registered and > have sufficient memory > 14/07/14 00:00:53 WARN TaskSchedulerImpl: Initial job has not accepted any > resources; check your cluster UI to ensure that workers are registered and > have sufficient memory > Training: 10 > 14/07/14 00:01:09 WARN BLAS: Failed to load implementation from: > com.github.fommil.netlib.NativeSystemBLAS > 14/07/14 00:01:09 WARN BLAS: Failed to load implementation from: > com.github.fommil.netlib.NativeRefBLAS > *Exception in thread "main" org.apache.spark.SparkException: Job aborted due > to stage failure: Serialized task 20:0 was 94453098 bytes which exceeds > spark.akka.frameSize (10485760 bytes). Consider using broadcast variables > for large values.* > at > org.apache.spark.scheduler.DAGScheduler.org$apache$spark$scheduler$DAGScheduler$$failJobAndIndependentStages(DAGScheduler.scala:1044) > at > org.apache.spark.scheduler.DAGScheduler$$anonfun$abortStage$1.apply(DAGScheduler.scala:1028) > at > org.apache.spark.scheduler.DAGScheduler$$anonfun$abortStage$1.apply(DAGScheduler.scala:1026) > at > scala.collection.mutable.ResizableArray$class.foreach(ResizableArray.scala:59) > at scala.collection.mutable.ArrayBuffer.foreach(ArrayBuffer.scala:47) > at > org.apache.spark.scheduler.DAGScheduler.abortStage(DAGScheduler.scala:1026) > at > org.apache.spark.scheduler.DAGScheduler$$anonfun$handleTaskSetFailed$1.apply(DAGScheduler.scala:634) > at > org.apache.spark.scheduler.DAGScheduler$$anonfun$handleTaskSetFailed$1.apply(DAGScheduler.scala:634) > at scala.Option.foreach(Option.scala:236) > at > org.apache.spark.scheduler.DAGScheduler.handleTaskSetFailed(DAGScheduler.scala:634) > at > org.apache.spark.scheduler.DAGSchedulerEventProcessActor$$anonfun$receive$2.applyOrElse(DAGScheduler.scala:1229) > at akka.actor.ActorCell.receiveMessage(ActorCell.scala:498) > at akka.actor.ActorCell.invoke(ActorCell.scala:456) > at akka.dispatch.Mailbox.processMailbox(Mailbox.scala:237) > at akka.dispatch.Mailbox.run(Mailbox.scala:219) > at > akka.dispatch.ForkJoinExecutorConfigurator$AkkaForkJoinTask.exec(AbstractDispatcher.scala:386) > at > scala.concurrent.forkjoin.ForkJoinTask.doExec(ForkJoinTask.java:260) > at > scala.concurrent.forkjoin.ForkJoinPool$WorkQueue.runTask(ForkJoinPool.java:1339) > at > scala.concurrent.forkjoin.ForkJoinPool.runWorker(ForkJoinPool.java:1979) > at > scala.concurrent.forkjoin.ForkJoinWorkerThread.run(ForkJoinWorkerThread.java:107) > > > > > -- > View this message in context: > http://apache-spark-user-list.1001560.n3.nabble.com/Error-when-testing-with-large-sparse-svm-tp9592.html > Sent from the Apache Spark User List mailing list archive at Nabble.com.