hadoop 没有用于方案的文件系统：文件

Question

I am trying to run a simple NaiveBayesClassifer using hadoop, getting this error

Exception in thread "main" java.io.IOException: No FileSystem for scheme: file
    at org.apache.hadoop.fs.FileSystem.createFileSystem(FileSystem.java:1375)
    at org.apache.hadoop.fs.FileSystem.access$200(FileSystem.java:66)
    at org.apache.hadoop.fs.FileSystem$Cache.get(FileSystem.java:1390)
    at org.apache.hadoop.fs.FileSystem.get(FileSystem.java:196)
    at org.apache.hadoop.fs.FileSystem.get(FileSystem.java:95)
    at org.apache.hadoop.fs.FileSystem.get(FileSystem.java:180)
    at org.apache.hadoop.fs.Path.getFileSystem(Path.java:175)
    at org.apache.mahout.classifier.naivebayes.NaiveBayesModel.materialize(NaiveBayesModel.java:100)

Code :

    Configuration configuration = new Configuration();
    NaiveBayesModel model = NaiveBayesModel.materialize(new Path(modelPath), configuration);// error in this line..

modelPath is pointing to NaiveBayes.bin file, and configuration object is printing - Configuration: core-default.xml, core-site.xml

I think its because of jars, any ideas?

Answer 1

This is a typical case of the maven-assembly plugin breaking things.

Why this happened to us

Different JARs ( hadoop-commons for LocalFileSystem , hadoop-hdfs for DistributedFileSystem ) each contain a different file called org.apache.hadoop.fs.FileSystem in their META-INFO/services directory. This file lists the canonical classnames of the filesystem implementations they want to declare (This is called a Service Provider Interface implemented via java.util.ServiceLoader , see org.apache.hadoop.FileSystem#loadFileSystems ).

When we use maven-assembly-plugin , it merges all our JARs into one, and all META-INFO/services/org.apache.hadoop.fs.FileSystem overwrite each-other. Only one of these files remains (the last one that was added). In this case, the FileSystem list from hadoop-commons overwrites the list from hadoop-hdfs , so DistributedFileSystem was no longer declared.

How we fixed it

After loading the Hadoop configuration, but just before doing anything FileSystem -related, we call this:

    hadoopConfig.set("fs.hdfs.impl", 
        org.apache.hadoop.hdfs.DistributedFileSystem.class.getName()
    );
    hadoopConfig.set("fs.file.impl",
        org.apache.hadoop.fs.LocalFileSystem.class.getName()
    );

Update: the correct fix

It has been brought to my attention by krookedking that there is a configuration-based way to make the maven-assembly use a merged version of all the FileSystem services declarations, check out his answer below.

Answer 2

For those using the shade plugin, following on david_p's advice, you can merge the services in the shaded jar by adding the ServicesResourceTransformer to the plugin config:

  <plugin>
    <groupId>org.apache.maven.plugins</groupId>
    <artifactId>maven-shade-plugin</artifactId>
    <version>2.3</version>
    <executions>
      <execution>
        <phase>package</phase>
        <goals>
          <goal>shade</goal>
        </goals>
        <configuration>
          <transformers>
            <transformer implementation="org.apache.maven.plugins.shade.resource.ServicesResourceTransformer"/>
          </transformers>
        </configuration>
      </execution>
    </executions>
  </plugin>

This will merge all the org.apache.hadoop.fs.FileSystem services in one file

Answer 3

For the record, this is still happening in hadoop 2.4.0. So frustrating...

I was able to follow the instructions in this link: http://grokbase.com/t/cloudera/scm-users/1288xszz7r/no-filesystem-for-scheme-hdfs

I added the following to my core-site.xml and it worked:

<property>
   <name>fs.file.impl</name>
   <value>org.apache.hadoop.fs.LocalFileSystem</value>
   <description>The FileSystem for file: uris.</description>
</property>

<property>
   <name>fs.hdfs.impl</name>
   <value>org.apache.hadoop.hdfs.DistributedFileSystem</value>
   <description>The FileSystem for hdfs: uris.</description>
</property>

Answer 4

Took me ages to figure it out with Spark 2.0.2, but here's my bit:

val sparkBuilder = SparkSession.builder
.appName("app_name")
.master("local")
// Various Params
.getOrCreate()

val hadoopConfig: Configuration = sparkBuilder.sparkContext.hadoopConfiguration

hadoopConfig.set("fs.hdfs.impl", classOf[org.apache.hadoop.hdfs.DistributedFileSystem].getName)

hadoopConfig.set("fs.file.impl", classOf[org.apache.hadoop.fs.LocalFileSystem].getName)

And the relevant parts of my build.sbt :

scalaVersion := "2.11.8"
libraryDependencies += "org.apache.spark" %% "spark-core" % "2.0.2"

I hope this can help!

Answer 5

thanks david_p,scala

conf.set("fs.hdfs.impl", classOf[org.apache.hadoop.hdfs.DistributedFileSystem].getName);
conf.set("fs.file.impl", classOf[org.apache.hadoop.fs.LocalFileSystem].getName);

or

<property>
 <name>fs.hdfs.impl</name>
 <value>org.apache.hadoop.hdfs.DistributedFileSystem</value>
</property>

Answer 6

For maven, just add the maven dependency for hadoop-hdfs (refer to the link below) will solve the issue.

http://mvnrepository.com/artifact/org.apache.hadoop/hadoop-hdfs/2.7.1

Answer 7

Assuming that you are using mvn and cloudera distribution of hadoop. I'm using cdh4.6 and adding these dependencies worked for me.I think you should check the versions of hadoop and mvn dependencies.

<dependency>
        <groupId>org.apache.hadoop</groupId>
        <artifactId>hadoop-core</artifactId>
        <version>2.0.0-mr1-cdh4.6.0</version>
    </dependency>

    <dependency>
        <groupId>org.apache.hadoop</groupId>
        <artifactId>hadoop-common</artifactId>
        <version>2.0.0-cdh4.6.0</version>
    </dependency>

    <dependency>
        <groupId>org.apache.hadoop</groupId>
        <artifactId>hadoop-client</artifactId>
        <version>2.0.0-cdh4.6.0</version>
    </dependency>

don't forget to add cloudera mvn repository.

<repository>
        <id>cloudera</id>
        <url>https://repository.cloudera.com/artifactory/cloudera-repos/</url>
</repository>

Answer 8

I use sbt assembly to package my project. I also meet this problem. My solution is here. Step1: add META-INF mergestrategy in your build.sbt

case PathList("META-INF", "MANIFEST.MF") => MergeStrategy.discard
case PathList("META-INF", ps @ _*) => MergeStrategy.first

Step2: add hadoop-hdfs lib to build.sbt

"org.apache.hadoop" % "hadoop-hdfs" % "2.4.0"

Step3: sbt clean; sbt assembly

Hope the above information can help you.

Answer 9

I assume you build sample using maven.

Please check content of the JAR you're trying to run. Especially META-INFO/services directory, file org.apache.hadoop.fs.FileSystem . There should be list of filsystem implementation classes. Check line org.apache.hadoop.hdfs.DistributedFileSystem is present in the list for HDFS and org.apache.hadoop.fs.LocalFileSystem for local file scheme.

If this is the case, you have to override referred resource during the build.

Other possibility is you simply don't have hadoop-hdfs.jar in your classpath but this has low probability. Usually if you have correct hadoop-client dependency it is not an option.

Answer 10

Another possible cause (though the OPs question doesn't itself suffer from this) is if you create a configuration instance that does not load the defaults:

Configuration config = new Configuration(false);

If you don't load the defaults then you won't get the default settings for things like the FileSystem implementations which leads to identical errors like this when trying to access HDFS. Switching to the parameterless constructor of passing in true to load defaults may resolve this.

Additionally if you are adding custom configuration locations (eg on the file system) to the Configuration object be careful of which overload of addResource() you use. For example if you use addResource(String) then Hadoop assumes that the string is a class path resource, if you need to specify a local file try the following:

File configFile = new File("example/config.xml");
config.addResource(new Path("file://" + configFile.getAbsolutePath()));

Answer 11

I faced the same problem. I found two solutions: (1) Editing the jar file manually:

Open the jar file with WinRar (or similar tools). Go to Meta-info > services , and edit "org.apache.hadoop.fs.FileSystem" by appending:

org.apache.hadoop.fs.LocalFileSystem

(2) Changing the order of my dependencies as follow

<dependencies>
<dependency>
  <groupId>org.apache.hadoop</groupId>
  <artifactId>hadoop-hdfs</artifactId>
  <version>3.2.1</version>
</dependency>

<dependency>
  <groupId>org.apache.hadoop</groupId>
  <artifactId>hadoop-common</artifactId>
  <version>3.2.1</version>
</dependency>

<dependency>
  <groupId>org.apache.hadoop</groupId>
  <artifactId>hadoop-mapreduce-client-core</artifactId>
  <version>3.2.1</version>
</dependency>

<dependency>
  <groupId>org.apache.hadoop</groupId>
  <artifactId>hadoop-client</artifactId>
  <version>3.2.1</version>
</dependency>



</dependencies>

Answer 12

It took me sometime to figure out fix from given answers, due to my newbieness. This is what I came up with, if anyone else needs help from the very beginning:

import org.apache.spark.SparkContext
import org.apache.spark.SparkConf

object MyObject {
  def main(args: Array[String]): Unit = {

    val mySparkConf = new SparkConf().setAppName("SparkApp").setMaster("local[*]").set("spark.executor.memory","5g");
    val sc = new SparkContext(mySparkConf)

    val conf = sc.hadoopConfiguration

    conf.set("fs.hdfs.impl", classOf[org.apache.hadoop.hdfs.DistributedFileSystem].getName)
    conf.set("fs.file.impl", classOf[org.apache.hadoop.fs.LocalFileSystem].getName)

I am using Spark 2.1

And I have this part in my build.sbt

assemblyMergeStrategy in assembly := {
  case PathList("META-INF", xs @ _*) => MergeStrategy.discard
  case x => MergeStrategy.first
}

Answer 13

Configuration conf = new Configuration();
conf.set("fs.defaultFS", "hdfs://nameNode:9000");
FileSystem fs = FileSystem.get(conf);

set fs.defaultFS works for me! Hadoop-2.8.1

Answer 14

For SBT use below mergeStrategy in build.sbt

mergeStrategy in assembly <<= (mergeStrategy in assembly) { (old) => {
    case PathList("META-INF", "services", "org.apache.hadoop.fs.FileSystem") => MergeStrategy.filterDistinctLines
    case s => old(s)
  }
}

Answer 15

This question is old, but I faced the same issue recently and the origin of the error was different than those of the answers here.

On my side, the root cause was due to hdfs trying to parse an authorithy when encountering // at the beginning of a path :

$ hdfs dfs -ls //dev
ls: No FileSystem for scheme: null

So try to look for a double slash or an empty variable in the path building part of your code.

Related Hadoop ticket: https://issues.apache.org/jira/browse/HADOOP-8087

Answer 16

Use this plugin

<plugin>
                <groupId>org.apache.maven.plugins</groupId>
                <artifactId>maven-shade-plugin</artifactId>
                <version>1.5</version>
                <executions>
                    <execution>
                        <phase>package</phase>
                        <goals>
                            <goal>shade</goal>
                        </goals>

                        <configuration>
                            <filters>
                                <filter>
                                    <artifact>*:*</artifact>
                                    <excludes>
                                        <exclude>META-INF/*.SF</exclude>
                                        <exclude>META-INF/*.DSA</exclude>
                                        <exclude>META-INF/*.RSA</exclude>
                                    </excludes>
                                </filter>
                            </filters>
                            <shadedArtifactAttached>true</shadedArtifactAttached>
                            <shadedClassifierName>allinone</shadedClassifierName>
                            <artifactSet>
                                <includes>
                                    <include>*:*</include>
                                </includes>
                            </artifactSet>
                            <transformers>
                                <transformer
                                    implementation="org.apache.maven.plugins.shade.resource.AppendingTransformer">
                                    <resource>reference.conf</resource>
                                </transformer>
                                <transformer
                                    implementation="org.apache.maven.plugins.shade.resource.ManifestResourceTransformer">
                                </transformer>
                                <transformer 
                                implementation="org.apache.maven.plugins.shade.resource.ServicesResourceTransformer">
                                </transformer>
                            </transformers>
                        </configuration>
                    </execution>
                </executions>
            </plugin>

Answer 17

If you are using sbt :

//hadoop
lazy val HADOOP_VERSION = "2.8.0"

lazy val dependenceList = Seq(

//hadoop
//The order is important: "hadoop-hdfs" and then "hadoop-common"
"org.apache.hadoop" % "hadoop-hdfs" % HADOOP_VERSION

,"org.apache.hadoop" % "hadoop-common" % HADOOP_VERSION
)

Answer 18

This is not related to Flink, but I've found this issue in Flink also.

For people using Flink, you need to download Pre-bundled Hadoop and put it inside /opt/flink/lib .

Answer 19

If you're using the Gradle Shadow plugin, then this is the config you have to add:

shadowJar {
    mergeServiceFiles()
}

Answer 20

I also came across similar issue. Added core-site.xml and hdfs-site.xml as resources of conf (object)

Configuration conf = new Configuration(true);    
conf.addResource(new Path("<path to>/core-site.xml"));
conf.addResource(new Path("<path to>/hdfs-site.xml"));

Also edited version conflicts in pom.xml. (eg If configured version of hadoop is 2.8.1, but in pom.xml file, dependancies has version 2.7.1, then change that to 2.8.1) Run Maven install again.

This solved error for me.