[HDFS-3846] Namenode deadlock in branch-1

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.1.0, 1.2.0
• Fix Version/s: 1.1.1
• Component/s: namenode
• Assignee: Brandon Li
• Reporter: Tsz-wo Sze
• Created: 23/Aug/12 09:09
• Updated: 16/Oct/13 09:45
• Resolved: 23/Aug/12 22:35

Description

Jitendra found out the following problem:
1. Handler : Acquires namesystem lock waits on SafemodeInfo lock at SafeModeInfo.isOn()
2. SafemodeMonitor : Calls SafeModeInfo.canLeave() which is synchronized so SafemodeInfo lock is acquired, but this method also causes following call sequence needEnter() -> getNumLiveDataNodes() -> getNumberOfDatanodes() -> getDatanodeListForReport() -> getDatanodeListForReport() . The getDatanodeListForReport is synchronized with FSNamesystem lock.

Issue Links

• relates to: HDFS-5368 Namenode deadlock during safemode extention