On 18-Aug-06, at 3:22 AM, Francisco Javier Cabello wrote:
Hello,
I have been 'googling' and I have found a lot of people warning
about the
problems with IDE write cache and journaling filesystems.
Should I disable write cache in my systems using reiserfs3+2.4.25?
I have tried to disable write cache with hdparm (hdparm -W0 /dev/
hdc) but it
is not working ( you can see write cache is enabled with hdparm
-i /dev/hdc).
Is there other way to disable write cache?
Get a UPS. :)
Here are some links I collected on this subject a few months ago.
* _Due to loose interpretations and vendor uniqueness in the ATA
Standard, there is no defined way that a driver can be assured that
the disk's cache has been flushed._ (http://developer.apple.com/
technotes/tn/tn1040.html)
* _if write back cache is turned on, it is not difficult to
create metadata inconsistency or corruption at the file system upon
power failure._ (http://sr5tech.com/write_back_cache_experiments.htm)
* Apple forum [[http://lists.apple.com/archives/darwin-dev/2005/
Feb/msg00072.html post]] discussing the issues and OS X's F_FULLFSYNC
feature, which tries hard to flush drive caches.
* Linux kernel mailing list: _How long can the unwritten data
linger in the drive cache if the drive is otherwise idle?_ http://
lkml.org/lkml/2003/11/2/73
* Interesting blog [[http://peter-zaitsev.livejournal.com/
12639.html?mode=reply post]] on the issue by a !MySQL developer.
_Transaction will be durable and database intact on the crash only if
database will perform synchronous IO as synchronous - reporting it is
done when data is physically on the disk._
* Detailed [[http://www.opensolaris.org/os/community/arc/caselog/
2004/652/ post]] about Open Solaris' approach to the issue (see
'spec' link)
Regards,
Paco
--
One of my most productive days was throwing away 1000 lines of code
(Ken
Thompson)
-----------------
PGP fingerprint: AF69 62B4 97EB F5BB 2C60 B802 568A E122 BBBE 5820
PGP Key available at http://pgp.mit.edu
-----------------
