Re: Two hour parity check

["Dom De Vitto" <dom@DeVitto.com>]

I think the softupdates "problems" were simply because the FAQ wasn't
updated for quite a while, and it said the code wasn't ready for
production (or something like that).

I run softupdates on all my writable filesystems, except mfs :-)

The speed increase on untarring and rm -rf'ing (which I do a lot of
as a admin/developer, YMMV) is dramatic.

Dom
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Dom De Vitto                                       Tel. 07855 805 271
http://www.devitto.com                         mailto:dom@devitto.com
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Where do you want to go today?  Same as every day.... Windows Update.
-----Original Message-----
From: owner-misc@openbsd.org [mailto:owner-misc@openbsd.org] On Behalf Of
Scott Francis
Sent: Friday, February 06, 2004 11:08 PM
To: Gary
Cc: misc@openbsd.org
Subject: Re: Two hour parity check

On Tue, Feb 03, 2004 at 03:36:26AM -0800, vex555@gmx.net said:
> On Mon, 2 Feb 2004 09:57:43 -0800
> Scott Francis <darkuncle@darkuncle.net> wrote:
>
> > be glad it's only 2 hours. :) I set up a RAID5 of 4x160GB 5400RPM 
> > IDE disks a while back (2 disks per promise PCI IDE card), and 
> > parity rebuild on that takes _fourteen_ hours. Not fun. However ...
>
> wow - thats totally useless for a web/mail server. Even two hours 
> every couple of months is going to have clients bitching.

don't crash the server and you won't have to do the parity rebuild. :)

(I know, I know ...)

> > definitely you want multiple filesystems. At the minimum, / /tmp 
> > /var /home/usr and /usr/local.
>
> Done this ecxept for the /usr/local. Why that one? Also I realise that 
> mail/web docs/ logs etc go in /var so that has to be the biggest, but 
> how big do /usr /usr/local /home and /tmp need to be ? (P3 1.2gig 1gig 
> ram, 2 gig swap)

take advantage of different mount flags for /usr and /usr/local -
theoretically, after you do the initial install, you could mount /usr
read-only (yes, you'd have to have /usr{obj,ports,local,X11R6,src} all
mounted separately, so probably not feasible) ... I guess mainly because
I've always split up /usr/local (where new software installs typically go)
from /usr. If an install blows up, it doesn't hose /usr. Since post-install
nothing is in /usr/local/{bin,sbin}, you can recover to post-install more
quickly losing /usr/local than losing all of /usr. *shrug*

> > fsck will run on every filesystem not cleanly unmounted. However, 
> > that probably won't be any slower than a single massive partition 
> > (and you may find that it will be faster). You may also wish to use 
> > the soft updates mount option on your RAID filesystem; I did that 
> > (450GB
> > filesystem) and read/write performance improved considerably.
>
> I've heard of problems with softupdates. I want reliability and uptime 
> over performance.

Perhaps somebody else can comment on softupdates problems; I haven't heard
of anything that is concrete enough to outweigh the performance benefits,
especially on slower hardware.

> > cheap IDE solution. I avoid the parity rebuild problem by being 
> > _very_ careful to cleanly unmount the filesystem and the RAID set.
>
> The box will be in a co-lo location 4 hours drive from here with ups 
> and diesel backup so it should be OK - but who knows?

indeed, you never know what may happen.
--
       Scott Francis | darkuncle(at)darkuncle(dot)net | 0x5537F527 "I gave
you the chance of aiding me willingly, but you have elected the way of
pain!" -- Saruman, speaking for sysadmins everywhere

[demime 0.98d removed an attachment of type application/pgp-signature]