Re: EMC crashes

[Will Shackleford <shackle-at-cme.nist.gov>]

Matthew King wrote:
> I am very much a EMC newbie, as well as a Linux newbie. What does one do
> when the system crashes and there's no choice but to cold boot it? I get all
> sorts of error messages (Aiee and such) and the machine won't boot.
> 
> I've reloaded the system from the BDI disc once and would rather not do it
> again, if I can help it - there's gotta be an easier way!
> 
> Thanks any help,
> 
> Matt King
> Columbus, GA
> 
> 
> 
> 

Here is my long list of things to try when the system
crashes, although after you read it you might decide
reinstalling the BDI is the best alternative after all.

After it crashes but before you hit the reboot/power switch
you can try these:


<CNTRL><ALT> and one of the keys <F1> through <F7>

All pressed simultaneously are used to switch between 7 virtual screens.

Ussually X windows runs on 7 and text only consoles run on the rest,
so for example <CNTRL><ALT><F1> should switch to a text console.

If you can't get a text console skip to the next "************"


If you can get to a text console it may be that X windows is hung.

If X starts at bootup and you normally use a graphical login
You can kill X and all of the programs running on it by switching
run levels.

Log into the text console as root and run

init 3

to switch to run-level 3 and therefore kill X


init 5

to switch back to run-level 5 and therefore restart X

If you normally use startx from a text console to start X, you
can try using <cntrl><alt><backspace> to kill X or
the command

killall xinit
or

killall -KILL xinit

******************

If you can't get a text console you can probably still do a softer
reboot using the magic sys request keys

These have to be enabled before hand both in the kernel configuration
and in /etc/sysctl.conf

<alt><sysRq>s
<alt><sysRq>u
<alt><sysRq>b

keys on the same line are pressed all at the same time. The
sysRq key is ussualy the same as the "Print Screen" key.

The entry in sysctl.conf to add is
kernel.sysrq=1

You also need to run "sysctl -p" or reboot after modifying sysclt.conf.


********************

If you have to hit power button  without doing a controlled shut down, 
it is likely your harddrive will be left in an inconsistant state.
If you boot linux with an out of sync harddrive it should automatically 
run fsck.
fsck is similar to  scandisk it can take a long time and prints lots of
cryptic messages, but the great majority of the time it cleans 
everything up and will eventually boot linux and everything will
be ok.


*****************


On the hopefully very rare occasions that all of the above has failed. 
You can try using the install disk as a rescue disk. I beleive the 
command is "rescue" typed at the first redhat install screen. This
ussually dumps you in a text console. You could try running fsck
with a variety of different options  to fix problems fsck doesn't fix in 
automatic mode. You can also use fdisk to check the partition table. 
Another thing that has happened to me a couple of times was to
modify the kernel and reboot without having run lilo or properly edited
lilo.conf. You can run lilo
from the rescue environment, just be sure to find the copy of lilo.conf 
that was on the harddrive and use lilo -C /<mountpoint>/etc/lilo.conf.
You can use the mount command to help figure out what mountpoint is
..

Look for something like /dev/hd?? or /dev/sd??  on /mntpoint


-- Will