OSEC

Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com
 
Subject: Re: large postfix installations?
From: Matthias Andree (madt.e-technik.uni-dortmund.de)
Date: Tue Apr 04 2000 - 15:54:29 CDT

Ian C.Sison <ian.sqsr.com.ph> writes:

> I'm probably missing specific behaviour here, but if a controller or
> hard disk caches writes, and fools the OS into believing that the data
> the OS asked the controller to write has been written to the disk
> [which is most probably the case whether the OS is in sync or async
> write mode], and power goes out, the same data corruption and
> integrity problems will exist.

If any layer confirms "successfully written to disk" state without the
data actually on disk, the system is broken.

More particularly: If the drive claims the data have been committed to
disk, it must be able to recover the data after every power loss
regardless of point in time or time until the power comes back on---if
that drive has back-up batteries or capacitors and logic to commit its
cache data to a FlashROM, has back-up batteries to commit the data to
magnetic disk, it does not matter. If it acknowledges successful data
write, it is expected that the drive does not lie about that. That's the
essence of sync()ing things. 

-- 
Matthias Andree

                Where do you think you're going today?