I\'m trying to use the spark-cassandra-connector via spark-shell on dataproc, however I am unable to connect to my cluster. It appears that there is a version mismatch since
Unfortunately, Hadoop's dependency on Guava 11 (which doesn't have the Futures.withFallback method mentioned) is a longstanding issue and indeed Hadoop 2.7.1 still depends on Guava 11.
Spark core uses Guava 14, as can be seen here but this is worked around by shading Guava inside the Spark assembly:
$ jar tf /usr/lib/spark/lib/spark-assembly.jar | grep concurrent.Futures
org/spark-project/guava/util/concurrent/Futures$1.class
org/spark-project/guava/util/concurrent/Futures$2.class
org/spark-project/guava/util/concurrent/Futures$3.class
org/spark-project/guava/util/concurrent/Futures$4.class
org/spark-project/guava/util/concurrent/Futures$5.class
org/spark-project/guava/util/concurrent/Futures$6.class
org/spark-project/guava/util/concurrent/Futures$ChainingListenableFuture$1.class
org/spark-project/guava/util/concurrent/Futures$ChainingListenableFuture.class
org/spark-project/guava/util/concurrent/Futures$CombinedFuture$1.class
org/spark-project/guava/util/concurrent/Futures$CombinedFuture$2.class
org/spark-project/guava/util/concurrent/Futures$CombinedFuture.class
org/spark-project/guava/util/concurrent/Futures$FallbackFuture$1$1.class
org/spark-project/guava/util/concurrent/Futures$FallbackFuture$1.class
org/spark-project/guava/util/concurrent/Futures$FallbackFuture.class
org/spark-project/guava/util/concurrent/Futures$FutureCombiner.class
org/spark-project/guava/util/concurrent/Futures$ImmediateCancelledFuture.class
org/spark-project/guava/util/concurrent/Futures$ImmediateFailedCheckedFuture.class
org/spark-project/guava/util/concurrent/Futures$ImmediateFailedFuture.class
org/spark-project/guava/util/concurrent/Futures$ImmediateFuture.class
org/spark-project/guava/util/concurrent/Futures$ImmediateSuccessfulCheckedFuture.class
org/spark-project/guava/util/concurrent/Futures$ImmediateSuccessfulFuture.class
org/spark-project/guava/util/concurrent/Futures$MappingCheckedFuture.class
org/spark-project/guava/util/concurrent/Futures.class
$ javap -cp /usr/lib/spark/lib/spark-assembly.jar org.spark-project.guava.util.concurrent.Futures
Compiled from "Futures.java"
public final class org.spark-project.guava.util.concurrent.Futures {
public static <V, X extends java.lang.Exception> org.spark-project.guava.util.concurrent.CheckedFuture<V, X> makeChecked(org.spark-project.guava.util.concurrent.ListenableFuture<V>, com.google.common.base.Function<java.lang.Exception, X>);
public static <V> org.spark-project.guava.util.concurrent.ListenableFuture<V> immediateFuture(V);
public static <V, X extends java.lang.Exception> org.spark-project.guava.util.concurrent.CheckedFuture<V, X> immediateCheckedFuture(V);
public static <V> org.spark-project.guava.util.concurrent.ListenableFuture<V> immediateFailedFuture(java.lang.Throwable);
public static <V> org.spark-project.guava.util.concurrent.ListenableFuture<V> immediateCancelledFuture();
public static <V, X extends java.lang.Exception> org.spark-project.guava.util.concurrent.CheckedFuture<V, X> immediateFailedCheckedFuture(X);
public static <V> org.spark-project.guava.util.concurrent.ListenableFuture<V> withFallback(org.spark-project.guava.util.concurrent.ListenableFuture<? extends V>, org.spark-project.guava.util.concurrent.FutureFallback<? extends V>);
public static <V> org.spark-project.guava.util.concurrent.ListenableFuture<V> withFallback(org.spark-project.guava.util.concurrent.ListenableFuture<? extends V>, org.spark-project.guava.util.concurrent.FutureFallback<? extends V>, java.util.concurrent.Executor);
public static <I, O> org.spark-project.guava.util.concurrent.ListenableFuture<O> transform(org.spark-project.guava.util.concurrent.ListenableFuture<I>, org.spark-project.guava.util.concurrent.AsyncFunction<? super I, ? extends O>);
public static <I, O> org.spark-project.guava.util.concurrent.ListenableFuture<O> transform(org.spark-project.guava.util.concurrent.ListenableFuture<I>, org.spark-project.guava.util.concurrent.AsyncFunction<? super I, ? extends O>, java.util.concurrent.Executor);
public static <I, O> org.spark-project.guava.util.concurrent.ListenableFuture<O> transform(org.spark-project.guava.util.concurrent.ListenableFuture<I>, com.google.common.base.Function<? super I, ? extends O>);
public static <I, O> org.spark-project.guava.util.concurrent.ListenableFuture<O> transform(org.spark-project.guava.util.concurrent.ListenableFuture<I>, com.google.common.base.Function<? super I, ? extends O>, java.util.concurrent.Executor);
public static <I, O> java.util.concurrent.Future<O> lazyTransform(java.util.concurrent.Future<I>, com.google.common.base.Function<? super I, ? extends O>);
public static <V> org.spark-project.guava.util.concurrent.ListenableFuture<V> dereference(org.spark-project.guava.util.concurrent.ListenableFuture<? extends org.spark-project.guava.util.concurrent.ListenableFuture<? extends V>>);
public static <V> org.spark-project.guava.util.concurrent.ListenableFuture<java.util.List<V>> allAsList(org.spark-project.guava.util.concurrent.ListenableFuture<? extends V>...);
public static <V> org.spark-project.guava.util.concurrent.ListenableFuture<java.util.List<V>> allAsList(java.lang.Iterable<? extends org.spark-project.guava.util.concurrent.ListenableFuture<? extends V>>);
public static <V> org.spark-project.guava.util.concurrent.ListenableFuture<java.util.List<V>> successfulAsList(org.spark-project.guava.util.concurrent.ListenableFuture<? extends V>...);
public static <V> org.spark-project.guava.util.concurrent.ListenableFuture<java.util.List<V>> successfulAsList(java.lang.Iterable<? extends org.spark-project.guava.util.concurrent.ListenableFuture<? extends V>>);
public static <V> void addCallback(org.spark-project.guava.util.concurrent.ListenableFuture<V>, org.spark-project.guava.util.concurrent.FutureCallback<? super V>);
public static <V> void addCallback(org.spark-project.guava.util.concurrent.ListenableFuture<V>, org.spark-project.guava.util.concurrent.FutureCallback<? super V>, java.util.concurrent.Executor);
public static <V, X extends java.lang.Exception> V get(java.util.concurrent.Future<V>, java.lang.Class<X>) throws X;
public static <V, X extends java.lang.Exception> V get(java.util.concurrent.Future<V>, long, java.util.concurrent.TimeUnit, java.lang.Class<X>) throws X;
public static <V> V getUnchecked(java.util.concurrent.Future<V>);
static {};
}
You can follow the instructions here https://arjon.es/2015/making-hadoop-2.6-spark-cassandra-driver-play-nice-together/ to also do shading yourself during compilation. With spark-shell you may be able to get away with some changes in spark.driver.extraClassPath
as mentioned here, though collisions may then continue to arise at various points.
In case you are using gradle to create a fat jar, you can use relocate
in your shadowJar
:
shadowJar{
zip64 true
mergeServiceFiles()
relocate 'com.google', 'hidden.google'
}
I had the same problem with running Spark in Eclipse after attempting to download Google GWT. Download the jar file and specify it in the Build Path in Eclipse or ClassPath anywhere else --> http://www.java2s.com/Code/Jar/g/Downloadguava150rc1jar.htm