How to debug boot problem efficiently, nash
in [RedHat]
|
From: Larry on 30 Sep 2007 21:58 Hi, I have problem booting - it fails when trying to mount root disk (on the init script line "mount -o defaults --ro -t ext3 /dev/root /sysroot"). Without getting into the specifics of this particular problem at this point, I would like to find out if there is a more efficient way to debug this kind of problem. Here is what I currently do: 1) Boot on CD, do "linux rescue" (and mount the boot and root disk from there) 2) Add some sleep or echo statments to the init script (in the initrd) and then recreate the initrd image (to see more clearly at what point it fails) 3) Try to boot and note where it panics 4) Try fixing this from "linux rescue" (e.g. fixing kernel boot arguments in grub.conf and fstab, etc.) 5) Try booting again and stare in amazement at the kernel panic. 6) Repeat from 4 ad nauseum. Is there any way to boot interactively or into a nash interpreter (even though it is not technically a shell), so that you can pinpoint what the problem is without going through the above cycle? Just to be able to display the available devices or the output of an "ls" command would go a long way. Any advice would be greatly appreciated. Thanks, Larry Ps. This is CentOS 4.4 on an older Adaptec raid (I2O). The exact error message at this time is: "Mounting root filesystem Kernel panic - not syncing: Attempted to kill init! mount: error 6 mountin"
From: J.O. Aho on 1 Oct 2007 00:46 Larry wrote: > Hi, > > I have problem booting - it fails when trying to mount root disk (on the > init script line "mount -o defaults --ro -t ext3 /dev/root /sysroot"). > Without getting into the specifics of this particular problem at this point, > I would like to find out if there is a more efficient way to debug this kind > of problem. Here is what I currently do: > > 1) Boot on CD, do "linux rescue" (and mount the boot and root disk from > there) > 2) Add some sleep or echo statments to the init script (in the initrd) and > then recreate the initrd image (to see more clearly at what point it fails) > 3) Try to boot and note where it panics > 4) Try fixing this from "linux rescue" (e.g. fixing kernel boot arguments in > grub.conf and fstab, etc.) > 5) Try booting again and stare in amazement at the kernel panic. > 6) Repeat from 4 ad nauseum. > > Is there any way to boot interactively Usually you kan hold down the 'i' key during the booting, do that when you see the sysinit script starts. By the way, why don't you check the log files? /var/log/dmesg should have early bootup errors. -- //Aho
From: Larry on 1 Oct 2007 14:45 >> Is there any way to boot interactively > > Usually you kan hold down the 'i' key during the booting, do that when you > see > the sysinit script starts. Tried that, but did not seem to do anything. > By the way, why don't you check the log files? /var/log/dmesg should have The panic happens when trying to mount the root disk. The only log file at this point would be in the ramdisk, since the root disk is not yet mounted. This is one reason I am trying to find if there is a way to boot interactively, so that I can view log files and examine the contents of the /proc directory, etc. to find out why it is panicing when trying to mount the root disk. Thanks, Larry
From: J.O. Aho on 2 Oct 2007 00:45 Larry wrote: >>> Is there any way to boot interactively >> Usually you kan hold down the 'i' key during the booting, do that when you >> see >> the sysinit script starts. > Tried that, but did not seem to do anything. YOu could look at console 12, in most distros that is the dmesg out put. >> By the way, why don't you check the log files? /var/log/dmesg should have > The panic happens when trying to mount the root disk. The only log file at > this point would be in the ramdisk, since the root disk is not yet mounted. > This is one reason I am trying to find if there is a way to boot > interactively, so that I can view log files and examine the contents of the > /proc directory, etc. to find out why it is panicing when trying to mount > the root disk. Interactivity don't come until the sysinit stage, may even be that you need to get to the point where the init level is selected. It will usually only limit your interaction to select if a service will be started or not. Some of the posts on the net suggest that the error 6 is indicating that is a driver problem, is the adaptec raid driver something that is part of the default CentOS kernel or something you had to add yourself? If it's the later, then I guess you missed to add it at latest kernel update. -- //Aho
From: Michael C. on 2 Oct 2007 00:58
["Followup-To:" header set to alt.os.linux.] On Tue, 02 Oct 2007 06:45:26 +0200, J.O. Aho <user(a)example.net> wrote: > YOu could look at console 12, in most distros that is the dmesg out > put. Please define most distros. It doesn't appear to be the case for Debian, and I've never heard of it before. Michael C. -- mjchappell(a)verizon.net http://mcsuper5.freeshell.org/ Action may not always bring happiness; but there is no happiness without action. -- Benjamin Disraeli |