Davide Benedetto created SPARK-36917:
----------------------------------------
Summary: java.lang.ClassCastException: cannot assign instance of
java.lang.invoke.SerializedLambda to field
org.apache.spark.rdd.MapPartitionsRDD.f of type scala.Function3 in instance of
org.apache.spark.rdd.MapPartitionsRDD
Key: SPARK-36917
URL: https://issues.apache.org/jira/browse/SPARK-36917
Project: Spark
Issue Type: Bug
Components: Spark Core, Spark Submit
Affects Versions: 3.1.2
Environment: Ubuntu 20
Spark3.1.2-hadoop3.2
Hadoop 3.1
Reporter: Davide Benedetto
My spark Job fails with this error:
org.apache.spark.SparkException: Job aborted due to stage failure: Task 0 in
stage 0.0 failed 4 times, most recent failure: Lost task 0.3 in stage 0.0 (TID
3) (davben-lubuntu executor 2): java.lang.ClassCastException: cannot assign
instance of java.lang.invoke.SerializedLambda to field
org.apache.spark.rdd.MapPartitionsRDD.f of type scala.Function3 in instance of
org.apache.spark.rdd.MapPartitionsRDD
My OS Linux Ubuntu 20 is in this way organized: I have two user: /home/davben
and /home/hadoop. Into hadoop user I have installed hadoop 3.1 and
spark-3.1.2-hadoop3.2. Both users refers to java-8-openjdk Java installation.
The Spark job is launched from user davben on eclipse IDE in this way:
I create the spark conf and the spark session
{code:java}
System.setProperty("hadoop.home.dir", "/home/hadoop/hadoop");
SparkConf sparkConf = new SparkConf()
.setAppName("simple")
.setMaster("yarn")
.set("spark.executor.memory", "1g")
.set("deploy.mode", "cluster")
.set("spark.yarn.stagingDir", "hdfs://localhost:9000/user/hadoop/")
.set("spark.hadoop.fs.defaultFS","hdfs://localhost:9000")
.set("spark.hadoop.yarn.resourcemanager.hostname","localhost")
.set("spark.hadoop.yarn.resourcemanager.scheduler.address","localhost:8030")
.set("spark.hadoop.yarn.resourcemanager.address ","localhost:8032")
.set("spark.hadoop.yarn.resourcemanager.webapp.address","localhost:8088")
.set("spark.hadoop.yarn.resourcemanager.admin.address","localhost:8083")
SparkSession spark =
SparkSession.builder().config(sparkConf).getOrCreate();{code}
Then I create a dataset with two entries:
{code:java}
List<Row> rows = new ArrayList<>();
rows.add(RowFactory.create("a", "b"));
rows.add(RowFactory.create("a", "a"));
StructType structType = new StructType();
structType = structType.add("edge_1", DataTypes.StringType, false);
structType = structType.add("edge_2", DataTypes.StringType, false);
ExpressionEncoder<Row> edgeEncoder = RowEncoder.apply(structType);
Dataset<Row> edge = spark.createDataset(rows, edgeEncoder);
{code}
Then I print the content of the current dataset edge
{code:java}
edge.show();
{code}
Then I perform a map transformation on edge that upper cases the values of the
two entries and return the result in edge2
{code:java}
Dataset<Row> edge2 = edge.map(new MyFunction2(), edgeEncoder);{code}
The following is the code of MyFunction2
{code:java}
public class MyFunction2 implements MapFunction<Row, Row>, scala.Serializable {
private static final long serialVersionUID = 1L;
@Override public Row call(Row v1) throws Exception {
String el1 = v1.get(0).toString().toUpperCase();
String el2 = v1.get(1).toString().toUpperCase();
return RowFactory.create(el1,el2);
}
}{code}
Finally I show the content of edge2
{code:java}
edge2.show();
{code}
I can confirm that, checking on the hadoop UI a localhost:8088, the job is
submitted correctly, and
what sounds strange is that the first show is returned correctly in my console,
but the second one fails returning the up mentioned error.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
