Rescue bootup and low-level maintenance (Was: X installation calamity solved. Thank you!)
On Sat, 23 Aug 1997, Craig Sanders wrote:
> 'single' still has many daemon processes running - you can do some
> repair or maintainence work in single mode but you can't safely do an
> fsck because the disks are mounted rw.
>
> some other tasks (like moving a filesystem from one partition to
> another) are also best done when there are no daemons running.
>
> e.g. moving /var or /var/spool to another disk or partition while
> daemons that use that filesystem are running (for example syslogd,
> sendmail, innd, squid, and many others) is not a good idea. The only
> way to be sure that you have correctly transferred all of the data is
> to be sure that nothing is running which can change it while it's being
> copied.
Thank you for the instructive explanation Craig. These things should be
written down in a place where it is easy to find for the panic-struck
newbie (somewhere on the debian website?)
Actually I find the best way to do all kinds of low-level (and/or
emergency) work to boot an entirely separate spare installation.
On a 2500 MB harddisk it is easy to spare 40 MB for that purpose. In that
space you can put a lot more tools than would ever fit on the rescue
diskette. I added the partition and the kernel on it to /etc/lilo.conf,
but in regard of the purpose of the rescue partition I also made a
seperate bootdiskette for booting it.
> > As a minor matter, typing ``linux emergency'' did not work. I was
> > able to get on. Even though I was supposedly root, however, I was
> > told that all my files were read-only, so I could not change.
>
> that is because the root filesystem was mounted read-only. this is not a
> bug, it's a feature.
>
> you should have seen a message on the screen (just before the root
> password prompt) which told you that the root partition was mounted
> read-only and that you could remount it read-write by typing:
>
> mount -n -o rw,remount /
>
> i find it's useful to also run:
>
> mount /proc
> open
> open
> open
>
> before doing anything else. the /proc filesystem is so that you can get
> PIDs of running tasks (so you can run "kill" or "killall" for example
> - which is useful sometimes), and the "open"s are so that you have a
> few spare virtual consoles to do stuff in if required. when booted in
> emergency mode, ^C doesn't work to kill processes. setting it with stty
> doesn't fix it either.
You have to install the package "open" to be able to use it. It is an
optional package:
</home/joost>$ dpkg -s open
Package: open
Status: install ok installed
Priority: optional
Section: admin
Installed-Size: 29
Maintainer: Dominik Kubla <dominik@debian.org>
Version: 1.4-1
Depends: libc5 (>= 5.4.0-0)
Conflicts: kbd (<< 0.92-1)
Description: start a program on a new virtual terminal (VT)
Open opens a new vt and runs a command on it. It can be used as a simple
way to start several console logins without having to type your passwd
on each VT in turn. open can be used as a simpler to use replacement for
the doshell(8) command.
.
open is similar in functionality to the AIX/RS6000 command of the same
name.
Joost
--
TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to
debian-user-request@lists.debian.org .
Trouble? e-mail to templin@bucknell.edu .
Reply to: