[HBASE-12751] Allow RowLock to be reader writer

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.3.0, 2.0.0
• Fix Version/s: 2.0.0
• Component/s: regionserver
• Hadoop Flags: Incompatible change, Reviewed

Release Note

Locks on row are now reader/writer rather than exclusive.

Moves sequenceid out of HRegion and into MVCC class; MVCC is now in charge. A WAL append is still stamped in same way (we pass MVCC context in a few places where we previously we did not).

MVCC methods cleaned up. Make a bit more sense now. Less of them.

Simplifies our update of MemStore/WAL. Now we update memstore AFTER we add to WAL (but before we sync). This fixes possible dataloss when two edits came in with same coordinates; we could order the edits in memstore differently to how they arrived in the WAL.

Marked as an incompatible change because it breaks Distributed Log Replay, a feature we'd determined already was unreliable and to be removed.

• Assignee: Michael Stack
• Reporter: Elliott Neil Clark
• Created: 23/Dec/14 17:00
• Updated: 17/Jun/22 18:44
• Resolved: 19/Nov/15 23:15

Description

Right now every write operation grabs a row lock. This is to prevent values from changing during a read modify write operation (increment or check and put). However it limits parallelism in several different scenarios.

If there are several puts to the same row but different columns or stores then this is very limiting.

If there are puts to the same column then mvcc number should ensure a consistent ordering. So locking is not needed.

However locking for check and put or increment is still needed.