[HDFS-4208] NameNode could be stuck in SafeMode due to never-created blocks

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 1.1.0
• Fix Version/s: 1.1.2
• Component/s: namenode
• Target Version/s: 1.1.2
• Hadoop Flags: Reviewed
• Assignee: Brandon Li
• Reporter: Brandon Li
• Created: 18/Nov/12 22:45
• Updated: 06/Mar/13 09:55
• Resolved: 27/Nov/12 01:50

Description

In one test case, NameNode allocated a block and then was killed before the client got the addBlock response. After NameNode restarted, it couldn't get out of SafeMode waiting for the block which was never created. In trunk, NameNode can get out of SafeMode since it only counts complete blocks. However branch-1 doesn't have the clear notion of under-constructioned-block in Namenode.

JIRA HDFS-4212 is to track the never-created-block issue and this JIRA is to fix NameNode in branch-1 so it can get out of SafeMode when never-created-block exists.

The proposed idea is for SafeMode not to count the zero-sized last block in an under-construction file as part of total blcok count.

Issue Links

• is related to: HDFS-4452 getAdditionalBlock() can create multiple blocks if the client times out and retries.
• relates to: HDFS-4212 NameNode can't differentiate between a never-created block and a block which is really missing