[HDFS-4748] MiniJournalCluster#restartJournalNode leaks resources, which causes sporadic test failures

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 2.1.0-beta
• Component/s: qjm, test
• Hadoop Flags: Reviewed
• Assignee: Chris Nauroth
• Reporter: Chris Nauroth
• Created: 24/Apr/13 23:15
• Updated: 27/Aug/13 22:07
• Resolved: 27/Apr/13 14:33

Description

MiniJournalCluster#restartJournalNode stops a JournalNode and then recreates a new one with the same configuration. However, it does not maintain a reference to the new JournalNode instance, so therefore it doesn't get stopped inside MiniJournalCluster#shutdown. The JournalNode holds a file lock on its underlying storage, so this can cause sporadic failures in tests like TestQuorumJournalManager.

Issue Links

• relates to: HDFS-3602 Enhancements to HDFS for Windows Server and Windows Azure development and runtime environments
• supercedes: HDFS-4276 HDFS tests have multiple failures on Windows due to file locking conflict