[HDFS-3718] Datanode won't shutdown because of runaway DataBlockScanner thread

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 2.0.1-alpha
• Fix Version/s: 0.23.3, 2.0.2-alpha
• Component/s: datanode
• Hadoop Flags: Reviewed
• Assignee: Kihwal Lee
• Reporter: Kihwal Lee
• Created: 24/Jul/12 21:02
• Updated: 19/Aug/13 06:44
• Resolved: 15/Aug/12 01:05

Description

Datanode sometimes does not shutdown because the block pool scanner thread keeps running. It prints out "Starting a new period" every five seconds, even after shutdown() is called. Somehow the interrupt is missed.

DataBlockScanner will also terminate if datanode.shouldRun is false, but in DataNode#shutdown, DataBlockScanner#shutdown() is invoked before it is being set to false.

Is there any reason why datanode.shouldRun is set to false later?