lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Mark Miller <>
Subject Re: [jira] Resolved: (LUCENE-1044) Behavior on hard power shutdown
Date Sun, 04 Nov 2007 00:08:47 GMT
Derby does not guarantee atomicity? Other java DB's? I thought they did, 
but perhaps not. You cannot rig a simple system with some sort of rename 

If a hard power shutdown interrupted, on the next start the system could 
ideally be left in a condition that uses the previous commit point, or 
recognize the latest commit failed, clean it up, and go back to the 
previous commit point. There must be ways to do it that don't rely on a 
buffer being fully flushed, but rather on atomic file system operations 
(of which there must be some that can be counted on in java?).

Chris Hostetter wrote:
> : Is the new autocommit=false code atomic (the new check point is successfully
> : made and moved to or its not)? If not I imagine it could be made to be without
> : too much work right?
> No matter what work we do in Java code to try and garuntee atomicity, the 
> JVM can't garuntee that File IO buffers are flushed unless the JVM is 
> shutdown cleanly, so i don't see how we could possibly make any claims
> of atomicity in the event of hard process (or OS) termination.
> i'd be happy to be proven wrong by someone who knows more about IO, 
> filesystems, and the JVM Specification.
> -Hoss
> ---------------------------------------------------------------------
> To unsubscribe, e-mail:
> For additional commands, e-mail:

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message