Hi there, I am trying to run Spark on Mesos using a Docker image as executor, as mentioned http://spark.apache.org/docs/latest/running-on-mesos.html#mesos-docker-support .
I built a docker image using the following Dockerfile (which is based on https://github.com/apache/spark/blob/master/docker/spark-mesos/Dockerfile): FROM mesosphere/mesos:0.25.0-0.2.70.ubuntu1404 # Update the base ubuntu image with dependencies needed for Spark RUN apt-get update && \ apt-get install -y python libnss3 openjdk-7-jre-headless curl RUN curl http://www.carfab.com/apachesoftware/spark/spark-1.6.0/spark-1.6.0-bin-hadoop2.6.tgz | tar -xzC /opt && \ ln -s /opt/spark-1.6.0-bin-hadoop2.6 /opt/spark ENV SPARK_HOME /opt/spark ENV MESOS_NATIVE_JAVA_LIBRARY /usr/local/lib/libmesos.so Then I successfully ran spark-shell via this docker command: docker run --rm -it --net=host <registry>/<image>:<tag> /opt/spark/bin/spark-shell --master mesos://<master_host>:5050 --conf <registry>/<image>:<tag> So far so good. Then I wanted to call sc.textFile to load a file from S3, but I was blocked by some issues which I couldn't figure out. I've read https://dzone.com/articles/uniting-spark-parquet-and-s3-as-an-alternative-to and http://blog.encomiabile.it/2015/10/29/apache-spark-amazon-s3-and-apache-mesos, learned that I need to add hadood-aws-2.7.1 and aws-java-sdk-2.7.4 into the executor and driver's classpaths, in order to access s3 files. So, I added following lines into Dockerfile and build a new image. RUN curl https://repo1.maven.org/maven2/com/amazonaws/aws-java-sdk/1.7.4/aws-java-sdk-1.7.4.jar -o /opt/spark/lib/aws-java-sdk-1.7.4.jar RUN curl http://central.maven.org/maven2/org/apache/hadoop/hadoop-aws/2.7.1/hadoop-aws-2.7.1.jar -o /opt/spark/lib/hadoop-aws-2.7.1.jar Then I started spark-shell again with below command: docker run --rm -it --net=host <registry>/<image>:<tag> /opt/spark/bin/spark-shell --master mesos://<master_host>:5050 --conf <registry>/<image>:<tag> --conf spark.executor.extraClassPath=/opt/spark/lib/hadoop-aws-2.7.1.jar:/opt/spark/lib/aws-java-sdk-1.7.4.jar --conf spark.driver.extraClassPath=/opt/spark/lib/hadoop-aws-2.7.1.jar:/opt/spark/lib/aws-java-sdk-1.7.4.jar But below command failed when I ran it in spark-shell: scala> sc.textFile("s3a://<bucket_name>/<file_name>").count() [Stage 0:> (0 + 2) / 2]16/01/26 23:05:23 WARN TaskSetManager: Lost task 0.0 in stage 0.0 (TID 0, ip-172-16-14-203.us-west-2.compute.internal): java.lang.RuntimeException: java.lang.ClassNotFoundException: Class org.apache.hadoop.fs.s3a.S3AFileSystem not found at org.apache.hadoop.conf.Configuration.getClass(Configuration.java:2074) at org.apache.hadoop.fs.FileSystem.getFileSystemClass(FileSystem.java:2578) at org.apache.hadoop.fs.FileSystem.createFileSystem(FileSystem.java:2591) at org.apache.hadoop.fs.FileSystem.access$200(FileSystem.java:91) at org.apache.hadoop.fs.FileSystem$Cache.getInternal(FileSystem.java:2630) at org.apache.hadoop.fs.FileSystem$Cache.get(FileSystem.java:2612) at org.apache.hadoop.fs.FileSystem.get(FileSystem.java:370) at org.apache.hadoop.fs.Path.getFileSystem(Path.java:296) at org.apache.hadoop.mapred.LineRecordReader.<init>(LineRecordReader.java:107) at org.apache.hadoop.mapred.TextInputFormat.getRecordReader(TextInputFormat.java:67) at org.apache.spark.rdd.HadoopRDD$$anon$1.<init>(HadoopRDD.scala:237) at org.apache.spark.rdd.HadoopRDD.compute(HadoopRDD.scala:208) at org.apache.spark.rdd.HadoopRDD.compute(HadoopRDD.scala:101) at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:306) at org.apache.spark.rdd.RDD.iterator(RDD.scala:270) at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:38) at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:306) at org.apache.spark.rdd.RDD.iterator(RDD.scala:270) at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:66) at org.apache.spark.scheduler.Task.run(Task.scala:89) at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:213) at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145) at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615) at java.lang.Thread.run(Thread.java:745) Caused by: java.lang.ClassNotFoundException: Class org.apache.hadoop.fs.s3a.S3AFileSystem not found at org.apache.hadoop.conf.Configuration.getClassByName(Configuration.java:1980) at org.apache.hadoop.conf.Configuration.getClass(Configuration.java:2072) ... 23 more I checked hadoop-aws-2.7.1.jar, the org.apache.hadoop.fs.s3a.S3AFileSystem class file is in it. I also checked the Environment page of driver's Web UI at 4040 port, both hadoop-aws-2.7.1.jar and aws-java-sdk-1.7.4.jar are in the Classpath Entries (system path). And following code ran fine in spark-shell: scala> val clazz = Class.forName("org.apache.hadoop.fs.s3a.S3AFileSystem") clazz: Class[_] = class org.apache.hadoop.fs.s3a.S3AFileSystem scala> clazz.getClassLoader() res2: ClassLoader = sun.misc.Launcher$AppClassLoader@770848b9 So, I am confused why the task failed with "java.lang.ClassNotFoundException" Exception? Is there something wrong in the command line options I used to start spark-shell, or in the docker image, or in the "s3a://" url? Or is something related to the Docker executor of Mesos? I studied a bit https://github.com/apache/spark/blob/branch-1.6/core/src/main/scala/org/apache/spark/scheduler/cluster/mesos/MesosSchedulerBackend.scala but didn't understand it well... Appreciate if anyone will shed some lights on me. Thanks, Mao Geng
