[HADOOP-1063] MiniDFSCluster exists a race condition that lead to data node resources are not properly released

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.12.0
• Fix Version/s: 0.13.0
• Component/s: test
• Release Note: Resolved race condition in shutting down MiniDFSCluster data node that prevented resources from being deallocated properly.
• Assignee: Hairong Kuang
• Reporter: Hairong Kuang
• Created: 05/Mar/07 20:23
• Updated: 02/May/13 02:29
• Resolved: 23/Mar/07 20:47

Description

In MiniDFSCluster, there is a possibility that a data node gets shutted down before it is constructed. This leads to the situation that the data node's resources are not properly released. In Cygwin I observe that the data node directory is still kept locked after a miniDFSCluster is shut down.

Issue Links

• is depended upon by: HADOOP-1047 TestReplication fails because DFS does not guarantee all the replicas are placed when a file is closed