(just to complete a doc i'm writing on trouble shooting and
emergency booting, my final experiments when even booting to
single user mode or run level 1 is not going to work.)
two even more minimal booting scenarios i've just tested
are booting with the "emergency" word appended to the boot
line, or "init=/bin/sh". so what's the difference? (and
feel free to fill in the inevitable holes here.)
emergency mode
--------- ----
Section 9.4 of the RH 9 Customizaation discusses booting
to emergency mode, by simply appending the word "emergency"
to the kernel line at boot time. if you try this, you'll be
prompted for the root password to actually go into maintenance
mode, at which point, you have a single shell on VC 1. so
what can you do here?
first, "mount" shows (at least in my case) only two actual
mounts (so this is a good scenario if the root filesystem is
good, but some of your others are toast):
# mount
/dev/hda1 on / type ext3 (rw)
/proc on /proc type proc (rw)
and that's it. but if you try to see what's running with
"ps", you get a complaint that /proc is not mounted, that you
should make sure you have a line for /proc in /etc/fstab, and
that you have to mount it manually with:
# mount -t proc /proc /proc
it's a bit odd that "mount" shows /proc is mounted, and
i *do* have such a line in /etc/fstab (as we all do, i'm sure),
but you still apparently have to mount it manually, so i do,
at which point "ps" shows me:
1 init
... about a dozen more ...
15 bash
55 ps
so in emergency mode, "init" was started, and i have the single
shell i can work at. fair enough.
in this mode, i can manually mount any of my additional
filesystems, but oddly, they don't show up as being mounted
with "mount" or "cat /etc/mtab". however, they *are* listed
in /proc/mounts. go figure.
oddities with the docs
-------- ---- --- ----
the Cust Guide claims that emergency mode is useful in that
init files are not loaded, and it can be used if init is corrupted
or not working. that's a bit odd, since init was definitely started
for me here, and i'm not about to trash /sbin/init to see what
happens. anyone know if emergency mode can handle a damaged
/sbin/init?
also, section 9.4 claims that the root filesystem is mounted
read-only, but that wasn't true in my case. anyone care to clarify
either of these points?
init=/bin/sh mode
------------ ----
an even more minimal boot scenario is to boot by appending
"init=/bin/sh", good if you've really trashed your /sbin/init
program, since this tells the kernel to start a shell instead
of the normal init program.
the differences here from emergency mode are that you don't
get prompted for the root password -- you just get dumped into
superuser maintenance mode.
you get all the same weirdnesses as with emergency mode,
as in having to mount /proc and so on. but once you do, running
"ps" shows:
1 sh
...
14 kjournald
47 ps
which just verifies that, yes indeed, "sh" was invoked as
process 1 instead of init, as one would expect. at that point,
this looks very much like emergency mode, at least superficially.
rescue mode
------ ----
both of the above still require a working root filesystem,
at the very least. if you don't even have *that*, you're pretty
much forced to boot from, say, the CD-ROM in rescue mode, and
fix things from there.
so, as i read it, in order of getting more desperate and needing
a more minimal configuration to get your system running,
* single-user/run level 1
* "emergency" mode
* "init=/bin/sh"
* rescue mode off of the CD
any clarifications anyone wants to add?
rday
p.s. i didn't get into using root/boot floppies or other
stuff like that, which would give you even more options.
--
redhat-list mailing list
unsubscribe mailto:[EMAIL PROTECTED]
https://www.redhat.com/mailman/listinfo/redhat-list
