Roberto C. Sanchez wrote:
mke2fs(8) (which is what you get when you run 'man mkfs.ext3'), states:
-m reserved-blocks-percentage
Specify the percentage of the filesystem blocks reserved for the
super-user. This avoids fragmentation, and allows root-owned
daemons, such as syslogd(8), to continue to function correctly
after non-privileged processes are prevented from writing to the
filesystem. The default percentage is 5%.
So, the default is 5%, and you can make it as small or as large as you
want. Incidentally, that is a *good* thing, since it prevents a rouge
unprivileged process from crashing the system by filling the disks with
crap.
I think I tried that option, too.
Of course, you are more than welcome to use reiserfs, JFS, or XFS.
However, reiserfs, from my understanding, has its own set of issues and
JFS and XFS are best left to "experts" or at least to people who know
what they are doing. For example, please read these two threads about
XFS:
http://oss.sgi.com/archives/xfs/2005-06/msg00155.html
http://lkml.org/lkml/2005/6/29/10
I recommend also
http://www-128.ibm.com/developerworks/aix/library/au-unix-reiserFS/index.html
http://www.debian-administration.org/articles/388
People had problems with reiserfs, but that was long ago. It is true
that traditionaly ext3 is most reliable, but now reiserfs has
data=journal option that seems to improve reliability, as it states, in
the case of disaster, but also reduces write speed.
I myself am using reiserfs on all both my machines. On my desktop
computer since 2004 (a few months xfs), and on router since feb-2006.
Router is always on, and last year I had at least 50 power failures, and
no single problem. I would like to mention that there is always writing
to the drive. Also, while I used xfs, there was one situation, but it
was repaired with fsck.
My friend has similar configuration. He had serious damage on power
failure which he mostly managed to recover from, but on that drive also
appeared bad sectors.
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
