[DERBY-692] improve post commit committed delete handling when system crashes

• Type: Improvement
• Status: Open
• Priority: Minor
• Resolution: Unresolved
• Affects Version/s: 10.0.2.0
• Component/s: Store
• Assignee: Unassigned
• Reporter: Mike Matrigali
• Created: 09/Nov/05 04:56
• Updated: 14/Feb/09 00:53

Description

Currently space reclamation of committed deleted rows is handled by queueing to an in memory list a queue of work
items when certain conditions are noticed. The usual condition is that the last row on a page has been deleted. If the
system crashes then this queue of work is lost. Work could be done during recovery to try to recover from this. Some
possibilities include:
1) modify redo of delete to queue post commit if appropriate
2) modify redo of abort of insert to queue post commit if appropriate
3) moditfy redo loop which finds page that needs no redo to check if all rows are delted and queue post commit if appropriate.

Issue Links

• relates to: DERBY-691 committed deleted row space reclamation may be missed if delete is actually an aborted insert.