[HDFS-14341] Weird handling of plus sign in paths in WebHDFS REST API

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 3.1.1
• Component/s: webhdfs
• Assignee: Unassigned
• Reporter: Stefaan Lippens
• Created: 06/Mar/19 15:04
• Updated: 07/Mar/19 08:51

Description

We're using Hadoop 3.1.1 at the moment and have an issue with the handling of paths that contain plus signs (generated by Kafka HDFS Connector).

For example, I created this example directory tmp/plus+plus

$ hadoop fs -ls tmp/plus+plus
Found 1 items
-rw-r--r--   3 stefaan supergroup       7079 2019-03-06 14:31 tmp/plus+plus/foo.txt

When trying to list this folder through WebHDFS the naive way:

$ curl 'http://hadoopname05:9870/webhdfs/v1/user/stefaan/tmp/plus+plus?user.name=stefaan&op=LISTSTATUS'
{"RemoteException":{"exception":"FileNotFoundException","javaClassName":"java.io.FileNotFoundException","message":"File /user/stefaan/tmp/plus plus does not exist."}}

Fair enough, the plus sign + is a special character in URLs, let's encode it as %2B:

$ curl 'http://hadoopname05:9870/webhdfs/v1/user/stefaan/tmp/plus%2Bplus?user.name=stefaan&op=LISTSTATUS'
{"RemoteException":{"exception":"FileNotFoundException","javaClassName":"java.io.FileNotFoundException","message":"File /user/stefaan/tmp/plus plus does not exist."}}

Doesn't work.
After some trial and error I found that I could get it working by encode the thing twice ("+" -> "%2B" -> "%252B"):

 curl 'http://hadoopname05:9870/webhdfs/v1/user/stefaan/tmp/plus%252Bplus?user.name=stefaan&op=LISTSTATUS'
{"FileStatuses":{"FileStatus":[
{"accessTime":1551882704527,"blockSize":134217728,"childrenNum":0,"fileId":314914,"group":"supergroup","length":7079,"modificationTime":1551882704655,"owner":"stefaan","pathSuffix":"foo.txt","permission":"644","replication":3,"storagePolicy":0,"type":"FILE"}
]}}

Seems like there is some double decoding going on in WebHDFS REST API.

I also tried with some other special characters like @ and =, and for these it seems to work both when encoding once (%40 and %3D respectively) and encoding twice (%2540 and %253D respectively)

Issue Links

• relates to: HDFS-14323 Distcp fails in Hadoop 3.x when 2.x source webhdfs url has special characters in hdfs file path