[HIVE-13975] Hive import table fails if there is no write access to the source location

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: Import/Export
• Assignee: Chaoyu Tang
• Reporter: Chaoyu Tang
• Created: 09/Jun/16 02:49
• Updated: 09/Jun/16 02:49

Description

It seems not right that a write permission is needed on the source side for import table because the CopyTask in import needs to create a staging directory under the imported source directory. For a user who does not have the write permission to the source directory, you will get error like following

Caused by: java.lang.RuntimeException: Cannot create staging directory 'hdfs://quickstart.cloudera:8020/user/hive/exp_t1/.hive-staging_hive_2016-05-26_16-38-29_453_8739265934924968327-1': Permission denied: user=test1, access=WRITE, inode="/user/hive/exp_t1":anonymous:supergroup:drwxrwxr-x
 ...
org.apache.hadoop.hdfs.DistributedFileSystem.mkdirsInternal(DistributedFileSystem.java:952)
	at org.apache.hadoop.hdfs.DistributedFileSystem.mkdirs(DistributedFileSystem.java:945)
	at org.apache.hadoop.fs.FileSystem.mkdirs(FileSystem.java:1856)
	at org.apache.hadoop.hive.common.FileUtils.mkdir(FileUtils.java:518)
	at org.apache.hadoop.hive.ql.Context.getStagingDir(Context.java:234)
	... 23 more

There are three tasks involved in import table, CopyTask, DDLTask and MoveTask. I wonder if the CopyTask is really needed?