[plug] Linux File Systems under Woody

[Michael Hunt]

On Thu, 2003-01-30 at 09:37, Craig Ringer wrote:
> > My reccomendation is boot on a small ext3 partition (easier to deal with
> > when using lilo/grub, fsck's in seconds when it happens) and raid0
> > stripe with reiserfs for the rest.  Put a large identical swap on each
> > disk and set priority to same in fstab (enables parallel swap similar to
> > a raid 0 stripe).
> 
> I'll agree with all of that up to RAID 0. Raid zero is /asking/ for 
> trouble. If either disk fails, you're likely to lose the data off both. 
> As a result a RAID 0 array will fail at the failure time of the 
> shortest-lived disk, so it'll tend to fail much earlier /on average/ 
> than an individual disk.
> 
> If you need RAID, use RAID 1 (mirroring) for nice safe data or RAID 5 
> for a compromise that while it eats some CPU time and slows down writes, 
> allows you to use 2/3 or more of your data storage and provides a 
> measure of protection against disk failures.
> 
> Craig Ringer

RAID in my situation is not really an option since one of the drives is
a refurbish/replacement IBM 60GB that went bung on me a while back and
I'm not a 100% sure I can completly trust it. (I haven't really tested
it hard since I got it back and I am only planning on putting
non-crticial backed up stuff on it, or stuff I can affrod to lose untill
I am satisfied about its integrity). If I did go RAID I would probably
use the promise raid controller I have on board which supports RAID 0/1.
However I don't have enough disks for this at the moment.

Looks like I will then follow the suggestion of a small ext3 /boot
partition and the rest using ReiserFS. Swap will be split accross the
two drives as suggested. I was thinking of making my non critical disk
XFS but given the feedback I think I will stay away from it at the
moment.

Pity about XFS. Given that it is a SGI thing I thought it would have
beem a lot more robust than this.

Michael Hunt

P.S. Thanks for all the input guys.