[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: ext3 commit time



On Sat, Feb 16, 2002 at 07:25:32PM +0100, Martin Wuertele wrote:
> Hi Daniel!
> 
> On Sat, 16 Feb 2002, Daniel Faller wrote:
> 
> > I am running the ext3 filesystem on a laptop. The default value for the 
> > commit time of ext3 seems to be 30 sec, which is too short for the HD to spin 
> > down. How can I increase this timeout ?
> > 
> > Has anyone experineces of bad impacts in increasing this timeout to several 
> > minutes ?
> > 
> > standard-kernel :	2.4.17
> > e2fsprogs  version:	1.25-1
>  
> i think newsforge has had an interview with one of the ext3 developers
> recently and he mentioned that this was one of the current problems with
> ext3. it does override any setting that prevents it from syncing the
> filsystem.

No, I believe the article stated he had an ugly patch that fixed the
frequent write problem, but that it had some unacceptable side effects
like making fsync() a noop, and therefore the patch is not published.

I recall reading an interesting interview with Theo de Radt (OpenBSD)
where he argued against journaled file systems, saying they had solved
issues of data integrity in a much safer way with better performance.
I don't know how much merit there is to his argument, but it'd be
interesting to see some comparison tests, including ones testing for
exceptional conditions and data integrity as well as pure performance.

-- 
Eric G. Miller <egm2@jps.net>



Reply to: