[HDFS-4516] Client crash after block allocation and NN switch before lease recovery for the same file can cause readers to fail forever

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 2.0.3-alpha, 3.0.0-alpha1
• Fix Version/s: 2.3.0
• Component/s: namenode
• Target Version/s: 2.3.0
• Hadoop Flags: Reviewed
• Assignee: Vinayakumar B
• Reporter: Uma Maheswara Rao G
• Created: 20/Feb/13 07:09
• Updated: 12/May/16 18:16
• Resolved: 20/Nov/13 15:10

Description

If client crashes just after allocating block( blocks not yet created in DNs) and NN also switched after this, then new Namenode will not know about locs.

Further details will be in comment.

Issue Links

• is related to: HDFS-7924 NameNode goes into infinite lease recovery
• is related to: HDFS-5443 Delete 0-sized block when deleting an under-construction file that is included in snapshot
• relates to: HDFS-3222 DFSInputStream#openInfo should not silently get the length as 0 when locations length is zero for last partial block.