[HDFS-12803] We should not lock FsNamesystem even we operate a sub directory, we should refinement the lock

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Duplicate
• Affects Version/s: 2.7.1, 3.0.0-alpha3
• Component/s: hdfs
• Assignee: Unassigned
• Reporter: Baolong Mao
• Created: 12/Nov/17 07:18
• Updated: 15/Nov/17 16:14
• Resolved: 15/Nov/17 16:14

Description

An example:

If a client is doing mkdir or delete a file, other client will wait for the FSNamesystem's lock to do some operation.

I think we have to refinement the lock. we can lock the parent inode only.

Issue Links

• duplicates: HDFS-5453 Support fine grain locking in FSNamesystem