Delete hdfs folder from java

问题:

In a java app running on an edge node, I need to delete a hdfs folder, if it exists. I need to do that before running a mapreduce job (with spark) that output in the folder.

I found I could use the method

org.apache.hadoop.fs.FileUtil.fullyDelete(new File(url))

However, I can only make it work with local folder (i.e. file url on the running computer). I tried to use something like:

url = "hdfs://hdfshost:port/the/folder/to/delete";

with hdfs://hdfshost:port being the hdfs namenode IPC. I use it for the mapreduce, so it is correct. However it doesn't do anything.

So, what url should I use, or is there another method?

Note: here is the simple project in question.

回答1:

I do it this way:

    Configuration conf = new Configuration();
    conf.set("fs.hdfs.impl",org.apache.hadoop.hdfs.DistributedFileSystem.class.getName());
    conf.set("fs.file.impl",org.apache.hadoop.fs.LocalFileSystem.class.getName());
    FileSystem  hdfs = FileSystem.get(URI.create("hdfs://<namenode-hostname>:<port>"), conf);
    hdfs.delete("/path/to/your/file", isRecursive);

you don't need hdfs://hdfshost:port/ in your file path

回答2:

This site works for me.

Just add the following codes in my WordCount program will do:

import org.apache.hadoop.fs.*;

...
Configuration conf = new Configuration();

Path output = new Path("/the/folder/to/delete");
FileSystem hdfs = FileSystem.get(conf);

// delete existing directory
if (hdfs.exists(output)) {
    hdfs.delete(output, true);
}

Job job = Job.getInstance(conf, "word count");
...

You do not need to add hdfs://hdfshost:port explicitly.