[H-GEN] partitioning again

[James McPherson]

(Note reply-to: being general at humbug.org.au vs James McPherson <jmcphers at laurel.ocs.mq.edu.au>)


David Jericho writes:
 > On Wed, May 12, 1999 at 01:02:12AM +1000, Robert Brockway wrote:
 > > (Note reply-to: being general at humbug.org.au vs Robert Brockway <robert at blake.humbug.org.au>)
 > > > For servers and the like, I'm still fairly religious with respect to 
 > > > paritioning. The last thing we want is someone DoS'ing the filesystem
 > > > and stopping a root login.
 > > So far I've been religios about it too :)  But if ide really is
 > > dynamically allocating stuff all over the disk then we are kidding ourselves :)
 > I don't think that is what aj(?) was trying to get at. We can't be sure 
 > of how the disk places it's data, whether it stripes across platters, or
 > uses the platters in a linear progression along the logical disk.
 > This is why some disks have faster average seek times, and others faster
 > raw throughput figures. With my simple understand of RAID arrays, this would
 > be a deciding factor in which drives you would finally buy.

actually, if you're going to be buying disks to fill a raid array with, you
shouldn't necessarily go with the fastest spindle speed - you have to look at
a number of different factors like what level of performance is acceptable
(how many reads and writes per second), what bandwidth you have to make use of
(that includes ethernet, scsi bus _and_ cpu) and how much reliability you
want. 

 > 
 > The reason why I do it, is so if /var/spool/mail fills up, it won't stop
 > the web server logging or users from sending email.

that and /var/{log,adm}/* too. If you're running a license server (fsck I
_hate_ them) and logging its output to a file then that will fill up _very_
quickly, espcially when you have lots of separate applications listed. 

Robert, I don't like the idea of having a single partition per disk if you
just have one disk - there are too many opportunities for grievous failure if
you get it wrong, and if you get an error when fscking then you'd stand a
better than even chance (imho) that your system would be rendered dangerously
unstable if not actually unusable. Of course, my perspective is skewed
somewhat by the fact that I'm working with oracle databases and a large user
population, but I think safety (paranoia even!) and established norms are
useful to adhere to. As a frinstance, a while ago we had to replace the disk
that our main mail partition was on because it was getting serious read/write
errors. If that had been our / partition we would have had to take the machine
down to poweroff - but instead we were able to merely turn off mail delivery
while we plugged in a new disk and restored from the previous night's backup
and then recovered what we could from the old disk. The benefit is obvious for 
systems with large user populations. For a home system, well, it depends I
guess on just how close to the edge you wish to live ;)


James C. McPherson
--
Unix Systems Administrator            Phone: +61.2.9850.9418
Office of Computing Services            Fax: +61.2.9850.7433
Macquarie University   NSW    2109     remove the extra char 
AUSTRALIA			       in the reply-to field

--
This is list (humbug) general handled by majordomo at lists.humbug.org.au .
Postings only from subscribed addresses of lists general or general-post.