[HBASE-812] Compaction needs little better skip algo

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 0.2.0, 0.18.0
• Fix Version/s: 0.2.1, 0.18.0
• Component/s: regionserver
• Assignee: Daniel Leffel
• Reporter: Billy Pearson
• Created: 10/Aug/08 15:09
• Updated: 13/Sep/08 23:22
• Resolved: 12/Aug/08 04:21

Description

Looking at this section of one of my compaction's we have 3 files to compact the new algo is working great in my test but I see this below often we are skipping 2 out of the 3 files and compacting 1 file. 1 file is kind of a wast might as well just copy the file my suggestion is if there is only 1 file left after the new algo skips then just go on to the next column and skip the last file also. This will help improve compaction times a little more.

2008-08-10 10:00:45,310 DEBUG org.apache.hadoop.hbase.regionserver.HStore: Compaction size of 1339600874/size: 4.6m, skipped 2, 4851776
2008-08-10 10:00:45,438 DEBUG org.apache.hadoop.hbase.regionserver.HStore: started compaction of 1 files into /hbase/webdata/compaction.dir/1339600874/size/mapfiles/8653208152776334891
2008-08-10 10:00:46,838 DEBUG org.apache.hadoop.hbase.regionserver.HStore: moving /hbase/webdata/compaction.dir/1339600874/size/mapfiles/8653208152776334891 to /hbase/webdata/1339600874/size/mapfiles/7539342470259528578
2008-08-10 10:00:47,166 DEBUG org.apache.hadoop.hbase.regionserver.HStore: Completed compaction of 1339600874/size store size is 4.6m