[PIG-5042] Pig with Tez requires a staging directory with submitter ownership

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 0.16.0
• Component/s: tez

Environment

Hadoop 2.7.2, Pig 0.16.0, Tez 0.8.4

• Assignee: Unassigned
• Reporter: Alexandre Linte
• Created: 13/Oct/16 08:17
• Updated: 24/Oct/16 06:39

Description

With Pig 0.15.0, I have been able to work using the Tez execution engine without issues. After upgrading Pig to 0.16.0, I haven't been able to submit my jobs.

Since the beginning, the tez-site.xml is configured as below:

<property>
  <description>String value to a file path. The location of the Tez libraries which will be localized for DAGs. This follows the following semantics</description>
  <name>tez.lib.uris</name>
  <value>hdfs:///Products/TEZ/tez-0.8.4.tar.gz</value>
</property>
<property>
  <description>String value. Specifies a directory where Tez can create temporary job artifacts.</description>
  <name>tez.staging-dir</name>
  <value>/Products/TEZ/staging/</value>
</property>

The staging directory was set up with 777 rights, including a sticky bit:

[root@spark01 ~]# hadoop fs -ls /Products/TEZ/
Found 3 items
drwxrwxrwt+  - hdfs hadoop          0 2014-11-27 10:31 /Products/TEZ/staging
-rw-r--r--   3 hdfs hadoop   46518370 2016-09-02 11:06 /Products/TEZ/tez-0.8.4.tar.gz

Pig 0.16.0 seems to need a staging directory with submitter directory, even if the staging directory has full access to "other" like my previous configuration.
This issue looks like a regression to me, compared to Pig 0.15.0. Can you check that point?