[PIG-4115] Timezone inconsistency in Pig Oozie action fails with "(org.apache.pig.builtin.ToDate2ARGS)[datetime]"

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 0.12.0
• Assignee: Rohini Palaniswamy
• Reporter: Michalis Kongtongk
• Created: 11/Aug/14 11:21
• Updated: 14/Oct/14 17:25

Description

Running a Pig Action in Oozie :

CentOS

• mv /etc/localtime /etc/localtime.mv #backup your current tz
• ln -sf /usr/share/zoneinfo/Europe/Zurich /etc/localtime # set tz to Europe/Zurich

• create a hdfs://tmp/file.txt in hdfs with content "1"
  In Pig do:
  A = load '/tmp/file.txt' as (a:chararray);
  B = foreach A generate *, ToDate('02/11/1940', 'dd/MM/yyyy') ;
  dump B;

In Oozie Pig Action,
produce the same script in in workflow.xml and execute
this is where it'll fail.

...
ERROR 0: Exception while executing [POUserFunc (Name: POUserFunc(org.apache.pig.builtin.ToDate2ARGS)[datetime] - scope-254 Operator Key: scope-254) children: null at []]: 
java.lang.IllegalArgumentException: Cannot parse "02/11/1940": Illegal instant due to time zone offset transition (Europe/Zurich) 
...

Since Oozie is using Pig as a library, I believe they should behave the same.

We notice this inconsistency, when the OS is set to
$ date +%Z # timezone name
CEST