[HBASE-28427] FNFE related to 'master:store' when moving archived hfiles to global archived dir

• Type: Improvement
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 2.6.0, 2.4.18, 3.0.0-beta-2, 2.5.9
• Component/s: master
• Labels: pull-request-available

Environment

Centos7
hbase-4.0.0-alpha-1-SNAPSHOT

• Hadoop Flags: Reviewed
• Assignee: guluo Assign to me
• Reporter: guluo
• Created: 07/Mar/24 15:35
• Updated: 19/Mar/24 20:30
• Resolved: 18/Mar/24 13:57

Description

Sometimes we encounter the following log.

2024-02-26T18:11:34,934 WARN  [master:store-Store-Compactor] region.MasterRegionFlusherAndCompactor: Failed to move archived hfiles from file:/root/hbase-4.0.0-alpha-1-SNAPSHOT/tmp/hbase/MasterData/archive/data/master/store/1595e783b53d99cd5eef43b6debb2682/state to global dir file:/root/hbase-4.0.0-alpha-1-SNAPSHOT/tmp/hbase/archive/data/master/store/1595e783b53d99cd5eef43b6debb2682/statejava.io.FileNotFoundException: File file:/root/hbase-4.0.0-alpha-1-SNAPSHOT/tmp/hbase/MasterData/archive/data/master/store/1595e783b53d99cd5eef43b6debb2682/state does not exist        at org.apache.hadoop.fs.RawLocalFileSystem.listStatus(RawLocalFileSystem.java:733) ~[hadoop-common-3.3.5.jar:?]        at org.apache.hadoop.fs.FileSystem.listStatus(FileSystem.java:2045) ~[hadoop-common-3.3.5.jar:?]        at org.apache.hadoop.fs.FileSystem.listStatus(FileSystem.java:2089) ~[hadoop-common-3.3.5.jar:?]        at org.apache.hadoop.fs.ChecksumFileSystem.listStatus(ChecksumFileSystem.java:970) ~[hadoop-common-3.3.5.jar:?]        at org.apache.hadoop.hbase.master.region.MasterRegionUtils.moveFilesUnderDir(MasterRegionUtils.java:42) ~[hbase-server-4.0.0-alpha-1-SNAPSHOT.jar:4.0.0-alpha-1-SNAPSHOT]        at org.apache.hadoop.hbase.master.region.MasterRegionFlusherAndCompactor.moveHFileToGlobalArchiveDir(MasterRegionFlusherAndCompactor.java:145) ~[hbase-server-4.0.0-alpha-1-SNAPSHOT.jar:4.0.0-alpha-1-SNAPSHOT]        at org.apache.hadoop.hbase.master.region.MasterRegionFlusherAndCompactor.compact(MasterRegionFlusherAndCompactor.java:157) ~[hbase-server-4.0.0-alpha-1-SNAPSHOT.jar:4.0.0-alpha-1-SNAPSHOT]        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149) ~[?:1.8.0_231]        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624) ~[?:1.8.0_231]        at java.lang.Thread.run(Thread.java:748) ~[?:1.8.0_231] 

Analysis:
Against 'master:store', Archived hfiles would first be placed in {hbase.rootdir}/MasterData/archive/ after compacting, and then moved to global archived dir.
'master:store' has 3 column families:proc,rs,state, if there is empty in one of column families,such as state, no files in this column families will be archived, and {hbase.rootdir}/MasterData/archive/data/master/store/1595e783b53d99cd5eef43b6debb2682/state would do not exist.
Finally, we will get FileNotFoundException when moving these archived hfiles to global archived dir.
 
Although，Current code can avoid this effecting by catching Exception, However it is maybe better to check the dir exists before moving.

Issue Links

• links to: GitHub Pull Request #5756