Jason Ryan said the following at 06/20/2012 04:54 PM :

I encountered this same issue at the start of the year: it is simply a matter of completing the installation and chrooting in to install grub to both your devices: http://jasonwryan.com/blog/2012/02/11/lvm/

The wiki seems strongly to urge one NOT to use grub, and the example given is for syslinux, so that's what I've been doing. Indeed. it says that the 2011.08.19 Arch Linux installer does not not support GRUB2. (I'm a bit confused about whether when "grub" is mentioned it means "grub2"; the wiki could be a lot clearer about that -- to be on the safe side, I avoided grub entirely and went with syslinux instead). 2011.08.19 is the what one gets if one downloads the current installer. According to the wiki, the step of copying the bootloader is supposed to occur after the reboot, and your page says that that should be possible. I did not realise it could be done before the reboot (since the wiki doesn't mention that possibility) so I have not tried that. I was planning to follow the new instruction placed into the wiki by Paul Dann and mentioned in his e-mail <2457829.5nCasJRlDv@leto> once I got past the reboot stage. So perhaps, having now discovered that the bootloader can be copied before the reboot, what I need to do is to go through the entire procedure again and this time copy the bootloader with /usr/sbin/syslinux-install_update -iam BEFORE the reboot. Incidentally, I have tried to boot off both the disks in the RAID1 array; they both fail with the same error (unable to find /dev/md0). Doc -- Web: http://www.sff.net/people/N7DR

Damjan said the following at 06/20/2012 04:21 PM :

...

# reboot

----

When I try to reboot, I receive the error message: ERROR: device /dev/md0 not found ERROR: unable to find root device /dev/md0

What do I have to do to get past this error?

Doc

PS I tried to go back to the very beginning and walk through the instructions again, but when I do that, when I reach this step: mdadm --create /dev/md0 --level=1 --raid-devices=2 /dev/sd[cd]3 I am now informed that /dev/sdc3 is busy or unavailable.

why are the sata devices sdc and sdd ?

I have a different OS on sda and sdb (which are also a RAID1 pair).

did you setup the mkinitcpio.conf as it says here: https://wiki.archlinux.org/index.php/Software_RAID_and_LVM#Configure_system

Yes. Doc -- Web: http://www.sff.net/people/N7DR

Paul Gideon Dann said the following at 06/21/2012 03:41 AM :

On Wednesday 20 Jun 2012 11:27:54 D. R. Evans wrote:

...

When I try to reboot, I receive the error message: ERROR: device /dev/md0 not found ERROR: unable to find root device /dev/md0

To me, this sounds like the RAID array is being given the wrong name, or the mdadm hook isn't being added to /etc/mkinitcpio.conf.

Extract from /etc/mkinitcpio.conf (sorry about any possible wrapping issue): MODULES="dm_mod" ... HOOKS="base udev mdadm_udev lvm2 autodetect pata scsi sata filesystems usbinput fsck"

When you get the above error message, are you dropped to a busybox shell? If so, can you do:

# ls /dev/md*

...to see if the array is being started at all?

I am dropped to some sort of primitive shell (the prompt says "rootfs"). As you suspect, ls /dev/md* reports no such file or device. So I think we have a clue and it looks like you are right that the RAID is not being started, although I don't know why. A couple more facts that may provide useful information: 1. I also did an ls /etc And I see that there are only three entries: fstab, mtab and udev. I don't know if that's reasonable. I naïvely expected to see a populated /etc (since presumably /etc/mkinicpio.conf has been read at this point), but perhaps that expectation was incorrect. 2. I checked that the RAID will start correctly if I assemble the array from within a different OS, and it does so. Doc -- Web: http://www.sff.net/people/N7DR

On 21/06/12 at 06:13pm, Paul Gideon Dann wrote:

On Thursday 21 Jun 2012 09:44:03 D. R. Evans wrote:

...

Extract from /etc/mkinitcpio.conf (sorry about any possible wrapping issue):

MODULES="dm_mod" ... HOOKS="base udev mdadm_udev lvm2 autodetect pata scsi sata filesystems usbinput fsck"

I have two RAID setups that work well for me. On both, I don't have anything in the MODULES line, and the following hooks:

HOOKS="base udev autodetect pata scsi sata mdadm lvm2 filesystems"

I hope this helps. I don't know anything about the mdadm_udev hook.

“Assembly via udev is also possible using the mdadm_udev hook. Upstream prefers this method of assembly.” Arch Wiki /J -- http://jasonwryan.com/ [GnuPG Key: B1BD4E40]

"Jérôme M. Berger" said the following at 06/21/2012 11:13 AM :

D. R. Evans wrote:

...

Paul Gideon Dann said the following at 06/21/2012 03:41 AM :

...

On Wednesday 20 Jun 2012 11:27:54 D. R. Evans wrote:

...

When I try to reboot, I receive the error message: ERROR: device /dev/md0 not found ERROR: unable to find root device /dev/md0 To me, this sounds like the RAID array is being given the wrong name, or the mdadm hook isn't being added to /etc/mkinitcpio.conf.

Extract from /etc/mkinitcpio.conf (sorry about any possible wrapping issue):

MODULES="dm_mod" ... Try adding "raid1" there...

Extract of /etc/mkinicpio.conf now reads: MODULES="dm_mod raid1" ... HOOKS="base udev mdadm_udev lvm2 autodetect pata scsi sata filesystems usbinput fsck" It made no difference, though :-( Doc -- Web: http://www.sff.net/people/N7DR

Chris Sakalis said the following at 06/21/2012 01:15 PM :

On Thu, Jun 21, 2012 at 9:21 PM, D. R. Evans wrote:

...

"Jérôme M. Berger" said the following at 06/21/2012 11:13 AM :

...

D. R. Evans wrote:

...

Paul Gideon Dann said the following at 06/21/2012 03:41 AM :

...

On Wednesday 20 Jun 2012 11:27:54 D. R. Evans wrote:

...

When I try to reboot, I receive the error message: ERROR: device /dev/md0 not found ERROR: unable to find root device /dev/md0 To me, this sounds like the RAID array is being given the wrong name, or the mdadm hook isn't being added to /etc/mkinitcpio.conf.

Extract from /etc/mkinitcpio.conf (sorry about any possible wrapping issue):

MODULES="dm_mod" ... Try adding "raid1" there...

Extract of /etc/mkinicpio.conf now reads: MODULES="dm_mod raid1" ... HOOKS="base udev mdadm_udev lvm2 autodetect pata scsi sata filesystems usbinput fsck"

It made no difference, though :-(

Doc

-- Web: http://www.sff.net/people/N7DR

Sorry for the stupid question, but do you rebuild the init ram image after changing your mkinitcpio.conf ?

I don't understand the question, so it's not at all stupid. I'm just following the instructions on the wiki. The wiki says to edit mkinitcpio.conf immediately prior to the reboot, so that's what I'm doing. It doesn't mention any need to rebuild an image (nor how to do it). Doc -- Web: http://www.sff.net/people/N7DR

Chris Sakalis said the following at 06/21/2012 01:57 PM :

On Thu, Jun 21, 2012 at 10:39 PM, D. R. Evans wrote:

...

...

Sorry for the stupid question, but do you rebuild the init ram image after changing your mkinitcpio.conf ?

I don't understand the question, so it's not at all stupid. I'm just following the instructions on the wiki. The wiki says to edit mkinitcpio.conf immediately prior to the reboot, so that's what I'm doing. It doesn't mention any need to rebuild an image (nor how to do it).

Doc

-- Web: http://www.sff.net/people/N7DR

On the wiki, it does not mention anything about rebuilding your initramfs, because mkinitcpio is called automatically by the installer after the configuration phase. However:

mkinitcpio.conf is read by the mkinitcpio utility in order to create the initial ramdisk image. Essentially, it's a bunch of stuff needed by the kernel before the root filesystem is made available. After you edit the mkinitcpio.conf file, you HAVE to rebuild[1] your initramfs image, in order for the changes to actually take effect. If you just edit the file and then reboot, nothing is done.

My turn to ask a stupid question, because I'm sure your explanation makes perfect sense to someone who understands this stuff, but I don't: my previous experience with installing Linux on RAID was with Ubuntu, for which I didn't need to mess with any of this. Are you saying that between these instructions: --- Add the dm_mod module to the MODULES list in /etc/mkinitcpio.conf. Add the mdadm_udev and lvm2 hooks to the HOOKS list in /etc/mkinitcpio.conf after udev. --- and the next step: --- Once it is complete you can safely reboot your machine: # reboot --- there should be some additional step(s)?

[1] - https://wiki.archlinux.org/index.php/Mkinitcpio#Image_creation_and_activatio...

That seems to say that I should run mkinitcpio, but that step isn't mentioned, even adumbratively, anywhere on https://wiki.archlinux.org/index.php/Software_RAID_and_LVM#RAID_installation Hopelessly, completely, lost and beginning to despair.... Doc -- Web: http://www.sff.net/people/N7DR

Chris Sakalis said the following at 06/21/2012 02:39 PM :

Note that, the arch installer automatically generates the image after the configuration is done (the penultimate step of the installer - "Configure System"). I guess that's why it is not explicitly mentioned in the wiki.

If you edit your mkinitcpio.conf *without* or *after* running this step, then you have to create a new image on your own. Otherwise, the installer takes care of it.

OK; I understand. Thank you. I'll go back and make sure that this is happening properly when I have the stomach for it (not today!). Doc -- Web: http://www.sff.net/people/N7DR

On Monday 02 Jul 2012 13:42:39 D. R. Evans wrote:

I have finally reached the point where the various /dev/md<n> devices mount during the reboot....

Now I get a large number of error messages of the form: init: failed to create pty - disabling logging for job and: could not load /lib/modules/3.4.4-2-ARCH/modules.dep no such file or directory on the console, and then the system appears to simply halt, without completing the reboot.

What are these messages trying to tell me, and how do I fix them so that the reboot can complete?

When you say "reboot", you're saying that these messages appear as ArchLinux is trying to boot (after grub), right? Not as it's shutting down? Could you list the messages that appear before the error messages, so that we can get an idea of what stage you've got to in the boot process? Like David, I also think this smells like an initramfs issue, and chances are that if you can chroot into your system somehow and run "mkinitcpio -p linux", it might help. Paul

2012/7/3 D. R. Evans :

Paul Gideon Dann said the following at 07/03/2012 02:35 AM :

...

On Monday 02 Jul 2012 13:42:39 D. R. Evans wrote:

...

I have finally reached the point where the various /dev/md<n> devices mount during the reboot....

Now I get a large number of error messages of the form: init: failed to create pty - disabling logging for job and: could not load /lib/modules/3.4.4-2-ARCH/modules.dep no such file or directory on the console, and then the system appears to simply halt, without completing the reboot.

[...]

...

Could you list the messages that appear before the error messages, so that we can get an idea of what stage you've got to in the boot process?

That's non-trivial... there's no way to capture them, so I would have to hope that XON/XOFF works (which it probably does) and write them down by hand. I'll do that when I can summon up the enthusiasm for it. Frankly, after more than a week trying to install Arch on RAID, I'm awfully close to giving up and going back to Ubuntu, which Just Worked, at least insofar as the installation was concerned.

In the meantime, as far as I recall, the messages said that the various RAID arrays were up and running, and then immediately I started getting a series of the "failed to create pty" messages. But it's quite likely that I've forgotten one or two informative messages that appeared before the errors started to occur.

Hello Doc, the messages sound a bit like a read-only root filesystem. I understand you've been going through a lot of work, but it seems like there has been so much done on this system, that it's very hard to troubleshoot now. One thing that isn't entirely clear to me (anymore); was this on a fresh install, or were you converting an existing Archlinux-install to a Raid configuration? In the case of fresh install; would it be an option to start over from scratch? I ask this, because then it would be a lot easier to list all the neccesary steps one by one and hopefully get it running as wished in the first place. I'm willing to dig up some spare HDD's to try and get Arch installed on a RAID1 array, right from the start, but only if that is what you're looking for. mvg, Guus

2012/7/3 D. R. Evans :

Guus Snijders said the following at 07/03/2012 01:57 PM :

...

In the case of fresh install; would it be an option to start over from scratch?

This was a fresh install. [...] I am wondering if the problem is something crazy like the fact that the RAID pair I'm using is on sdc and sdd, and something somewhere is assuming that I'm using sda and/or sdb (which actually are a RAID1 pair with Kubuntu on them).

I'm thinking about simply pulling the plug on sda and sdb, and performing a completely fresh install again so that there's no chance whatsoever that at any point in the process can it go to sda or sdb by mistake.

Hmm, i thought that /etc/mdadm.conf should take care of that.

...

I ask this, because then it would be a lot easier to list all the neccesary steps one by one and hopefully get it running as wished in the first place. I'm willing to dig up some spare HDD's to try and get Arch installed on a RAID1 array, right from the start, but only if that is what you're looking for.

That is indeed what I'm looking for.

Let me try one last time; I don't want to drag any one else into the morass that I'm currently experiencing unless it's absolutely necessary. I just need to get my enthusiasm up to the point where I'm willing to spend another hour or two trying it again.

Ok. I guess you'd best contact me off-list if we are going side-by-side. If/when we find something wrong in the wiki we could report back here. mvg, Guus