Rebuilding World Problems

Kevin Oberman oberman at es.net
Thu Feb 14 04:02:49 UTC 2008


> Date: Wed, 13 Feb 2008 13:25:04 -0800
> From: "Chris H." <chris#@1command.com>
> Sender: owner-freebsd-stable at freebsd.org
> 
> Quoting Gavin Spomer <spomerg at cwu.EDU>:
> > Thanks, this was very helpful. First of all I would just like to 
> > admit that I only gave /usr/src/UPDATING a quick glance. Shame on me; 
> > it might have solved most (if not all) of my problem if I had read 
> > more thoroughly.
> >
> > Surprisingly I grokked most of your cheatsheet and looked at a few 
> > man pages to figure out most of the rest. Haven't tried it all yet. I 
> > was wondering about the "mount -u /". Is it really necessary to mount 
> > the root partition prior to mounting all of them in the next step?
> 
> Absolutely. Think about it for a momment. Given that EVERYTHING
> (save swap) is mounted off of root ( / ). So it becomes quite
> impossible to mount /usr/ if / hasn't already been mounted. In
> other words; if / hasn't been mounted it doesn't exist for usr/
> to mount from it. :)

Absolutely not. You are stretching the logic a bit too far on this. I'll
admit I am baffled by why imp added this to UPDATING.

If you have a valid fstab file, it will have a line (usually the first
non-comment line) that specifies the partition as (1) ufs and (2) rw. As
a result, 'mount -a -t ufs' will remount / as read-write before trying
to mount any other file systems.

In over a decade of using FreeBSD (since early V3 days), I have never
bothered to specify the explicit remount of /. 

> >
> > I don't really understand the "swapon -a". When is it necessary and 
> > when is it not?
> 
> As a rule, it is already available after boot. So executing swapon -a
> is often considered overkill. /But/ absolutely no harm will come of
> doing it, and it /may/ be necessary. So this just insures you have
> an "event free" journey. :)

Actually, swap is not enabled in single-user mode. It is enabled very
early in the startup sequence going to multiuser, but, for obvious
reasons, it can't be enabled until dumpon has run. It also starts after 
initrandom, geli, gbde, encswap, and ccd, although most people don't
have all of these enabled.

If you are upgrading a system with limited memory and don't start swap,
you may run out of RAM and the upgrade will freeze. Not good. Even if it
odes not freeze, memory fragmentation could significantly slow progress.

In most systems the installworld and mergemaster will never touch swap
and the step has no impact, but it never hurts.

> >
> > Also, UPDATING has "adjkerntz -i" just before "mergemaster -p". I 
> > looked at the man page for adjkerntz and am still uncertain if I need 
> > to do this. I run an ntpd client, if that makes any difference.
> 
> Again, just a precaution. Think "safe", or "event free". :)

Not at all. Many systems run with a hardware (BIOS) clock set to local
time. Your system will run with this time until ntpd can reset it near
the end of the init sequence. Many files may be created with broken
timestamps during this interval.

You can easily check by entering the command 'date' after the system
reaches the single-user prompt. If the time printed is correct, there is
no need for the 'adjkerntz -i'. If the time is an hour or more off, it
is needed. If you live in the UK or any other country in the 0 offset
time zone (not many) and it's not summer time, this is not an issue.

> >
> > I think the documentation is an excellent reference for people who 
> > already, moderately know FreeBSD. I am not even a true newbie as I 
> > have a CS degree and have been a Linux admin for 2 years. Even so I 
> > often have a hard time with the complexity of FreeBSD. I recognize 
> > the value of understanding the fine-grain "nuts and bolts" of a 
> > system, but even so I wonder if FreeBSD over-complicates some things?
> 
> This is the "UNIX way". It breaks everything into small bits of
> useful stuff. There-by providing the "nuts & bolts" to build, or
> accomplish almost /anything/ with little, or no effort. Linux kind
> of "missed the boat" on this one. But even Linus T. indicates that
> Linux is not UNIX. I'd have to say, it's more a "feels like UNIX"
> than anything else.

Sorry, but this is not Linus' doing. for better or worse. He provides a
kernel, not an operating system. Red Hat, Suse, Mandriva, Ubuntu,
Debian, et. al. provide operating systems that use Linus' kernel and
use the name Linux.
> 
> To sum things up; given that I've been using BSD since long before
> FreeBSD even existed. I can't imagine how anyone would consider
> using anything /but/ *BSD. It is /infinitely/ flexible, which only
> adds to it's power. While - as you mentioned, it seems complicated
> to a new user. One must remember, after all, that it is a /server/
> and perhaps, not best suited to an average "desktop" user. But, if
> given the time, will become your best friend - /really/. :)

I, too, have been using BSD for a bit longer than FreeBSD has existed,
having used it while supporting the UC Davis Department of Applied
Sciences back in the 1970s. Not that the CSRG BSD days are relevant to
much of this as the boot-up as well as the rebuild procedures have been
totally re-worked since then (thank goodness).
-- 
R. Kevin Oberman, Network Engineer
Energy Sciences Network (ESnet)
Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab)
E-mail: oberman at es.net			Phone: +1 510 486-8634
Key fingerprint:059B 2DDF 031C 9BA3 14A4  EADA 927D EBB3 987B 3751
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 224 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-stable/attachments/20080214/96ee07f7/attachment.pgp


More information about the freebsd-stable mailing list