[MAPREDUCE-2042] InputFormat#getSplits() is called twice in local mode

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Abandoned
• Affects Version/s: 0.20.1, 0.20.2
• Fix Version/s: 0.20.3
• Component/s: job submission
• Assignee: Unassigned
• Reporter: Johannes Zillmann
• Created: 28/Aug/10 21:07
• Updated: 10/Mar/15 02:58
• Resolved: 10/Mar/15 02:58

Description

In local mode the InputFormat#getSplits() is called twice in local mode.

• 1st time: JobClient#writeOld/NewSplits() (then they write the splits to disk )
• 2nd time: LocalJobRunner#run() (instead of reading the split file )

That can become annoying in case the InputFormat access external resources or takes a little longer to create the splits.