[HBASE-4992] Use a UUID for HFileOutputFormat partition path name

• Type: Bug
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 0.92.0
• Component/s: mapreduce
• Hadoop Flags: Reviewed
• Assignee: Unassigned
• Reporter: Scott Chen
• Created: 09/Dec/11 17:09
• Updated: 20/Nov/15 11:53
• Resolved: 09/Dec/11 19:08

Description

HFileOutputFormat uses timestamps for the path name to avoid collision.
It is possible the path name still collide and causing MR job fails.

Caused by: java.io.IOException: File: hdfs://ubasedev001-snc3-dfs-nn.data.facebook.com:9000/user/hadoop/partitions_1316142976586#_\
partition.lst has changed on HDFS since job started
at org.apache.hadoop.filecache.DistributedCache.ifExistsAndFresh(DistributedCache.java:572)
at org.apache.hadoop.filecache.DistributedCache.checkCacheStatusValidity(DistributedCache.java:465)
at org.apache.hadoop.filecache.DistributedCache.getLocalCache(DistributedCache.java:230)
at org.apache.hadoop.mapred.TaskRunner.run(TaskRunner.java:208)