[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Partitioning a Web Server



On Fri, 4 Apr 2003 09:29, Jones, Steven wrote:
> um....compaq dl320s for 1 wont do it. Look at some bioses, i would be

Compaq is history.

> pleased if you could point me at some machines that can, ive not found one

A cheap clone motherboard that I bought from a local computer fair three years 
ago supported it.  It even would give a menu on boot to select which device 
to boot from if you pressed certain keys.

> yet. Maybe you should actually try it? have you? so far Ive not found a box

Yes I have tried it.  I didn't get it fully working however because this was 
before the recent improvements in the LILO code for software RAID.  By the 
time the newer versions of LILO were around I had lost interest in it.

> Have you tested this on the Dell? sounds scsi, might be possible if the

Of course I have tested it on the Dell.  I even published a little HTML file 
on my experiences with the Dell (to this list too).  You must be new to this 
list.

> second device which was sdb now becomes sda,

I believe that this is standard SCSI functionality.  However I usually use IDE 
so I'm not an expert on this.

> Of the two machines with bioses that allowed alternative devices it would
> not boot. It would take alterations to grub or lilo, from a maintenance
> point of view its a non-starter.

Which version of LILO did you try?

After I get my next desktop machine I'll do some more tests on this and write 
some detailed docs.

> re, setting up lilo, this means on your 2 disks there is a seperate mbr
> setup etc...never seen a doc on this, if we are talking doing it each time

The standard setup for LILO on software RAID has the LILO blocks installed on 
both disks and the debian-mbr package used for the MBR.  If the BIOS makes 
the boot disk device 0x80 then all will be fine.  If not then it will be much 
more difficult (but should still be possible if you adjust the lilo config).

> we update the kernel its beginning to get messy and time consuming, its
> certainly not standard as far as I am aware. A hardware raid card is a way
> more reliable solution as its simpler.

If you are talking about IDE RAID then avoid all the cheap cards.  Promise etc 
just cause problems.  Your choices for IDE are a good card (the only option I 
know of is 3ware) or software RAID.

> As for testing the /var issue you suggest, I would need to come up with a
> test that covers all eventualities, rather difficult to guarantee, way
> better to assume the worst and plan accordingly.

In that case you would need separate file systems for /var/tmp, /var/log, 
/var/cache, /var/spool, /var/lib, and /var/lock.  :-#

In the course of SE Linux policy development I have tested denying many 
different file accesses for many different programs, and found no evidence to 
suggest that having /var become full will deny logins.

Before writing my previous email I deliberately filled /var and logged in as a 
non-root user with no problems.  I have also had /var fill up from cron jobs 
on many occasions without any such problems.

-- 
http://www.coker.com.au/selinux/   My NSA Security Enhanced Linux packages
http://www.coker.com.au/bonnie++/  Bonnie++ hard drive benchmark
http://www.coker.com.au/postal/    Postal SMTP/POP benchmark
http://www.coker.com.au/~russell/  My home page



Reply to: