[CRUNCH-502] OutputFormat has inconsistent context state in interface functions

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.12.0
• Fix Version/s: 0.12.0
• Component/s: IO
• Assignee: Josh Wills
• Reporter: Mārtiņš Kalvāns
• Created: 02/Mar/15 15:59
• Updated: 18/May/15 19:04
• Resolved: 03/Mar/15 21:55

Description

I created example project to demonstrate problematic behaviour:
https://github.com/sisidra/crunch-ofb

1. FormatBundle config is not populated to Configuration in checkOutputSpecs:
https://github.com/sisidra/crunch-ofb/blob/master/src/main/java/com/spotify/crunch/bugreport/MyOutputFormat.java#L39

15/03/02 15:40:24 INFO bugreport.MyOutputFormat: my.config.key (checkOutputSpecs): null
15/03/02 15:40:24 ERROR bugreport.MyOutputFormat: Wrong my.config.key value in checkOutputSpecs!

2. TaskAttemptContext. getTaskAttemptID().toString() is different in getRecordWriter and getOutputCommitter:

2015-03-02 15:40:38,960 INFO [main] com.spotify.crunch.bugreport.MyOutputFormat: TaskAttemptID (getOutputCommitter): attempt_1422406067005_0121_m_000000_0
...
2015-03-02 15:40:39,789 INFO [main] com.spotify.crunch.bugreport.MyOutputFormat: TaskAttemptID (getRecordWriter): attempt_1422406067005_out0_0121_m_000000_0