Hello everyone,

I have been banging my head against the brick wall which is the Debian installer.

I am trying to install to a system with 5 hard drives:

1) Boot drive on an Adaptec 29160
2) 2 x 120 GB on a vanilla Promise ATA 133 controller (Linux software RAID)
3) 2 x 250 GB on the motherboard's Intel ICH5 SATA controller (Linux software RAID)

Here's the issue:

The BIOS orders everything like it should:

IDE, SCSI, SATA

I then disable booting from IDE in the BIOS, and change it to SCSI. The BIOS on this board will actually identify each hard drive separately, so you can choose which to boot from.

The Debian installer, however, for some asinine reason identifies the drives in a completely backwards order!

So my SCSI drive which should be /dev/sda is labelled as /dev/sdc in the installer!!

Red Hat 9 and all derivatives thereof actually have an option when configuring the boot loader to re-order the drives, so you can put /dev/hda ahead of /dev/sda, thereby changing your device.map for you.

How can I do this with Debian during install??

I came over to Debian a few months ago from the Red Hat world after installing Ubuntu on my latop, falling in love with it, then building a couple of servers with Debian.

This whole process is frustrating me, and turning me off *very* quickly.

Does anyone know how I can work around this issue? I don't want to go back to RPM hell...

Thanks in advance,

- Lance

Hostile,

If I understand well, you have one SCSI drive which you want to boot from, connected to the Adaptec. You expect this to be /dev/sda.

Then you have two SATA disks connected, and you expect these to be /dev/sdb and /dev/sdc.

During installation, Debian insists dat the SCSI/Adapted is /dev/sdc

Correct?

What happens if you mount de root dir on /dev/sdc (as Debian proposes) and actually boot from /dev/sdc? Is it not possible to make your BIOS doing that?

Another alternative:

Install your root partition on /dev/sdc (The SCSI adaptec)
Use either the Debian installer or grub-install afterwards to install an MBR on any of the disks that the BIOS wants to boot, for example /dev/sda

Some clarification why this would be no problem.
The space of the MBR is not being used by any disk.
You can install the actual boot loader (that is everything you need to boot, minus the MBR) on any disk, on any partition. (for example /dev/sdc1/boot)
You can install the root dir on any disk on any partition, just as you like it.

Does that help?

jlinkels

Correct. The BIOS id's the Adaptec first, and the SATA controller second, so why Debian is pushing it to the back of the order baffles me.

Correct.

Correct.

If I install to /dev/sdc1, grub fails.

I make a habit of never installing to the MBR, always /dev/(x)da1. I have tried what you have suggested anyways, and has not worked thus far.

With how I am laying out this system, for the purposes of imaging and easy replacement of the boot drive should it fail, I wish everything neatly contained on /dev/sda, and to boot from /dev/sda1.

After having done some more research, the identification order of SCSI devices is left up to libata. libata is what is completely screwing people up with a mix of both SCSI and SATA devices in their systems. It identifies both as SCSI devices, and gives no means to arrange the order prior to boot.



At least any RH-based variant has the option when installing the boot loader to change the drive order, something Debian lacks.

Ok,

I recognize the feeling. "There might be a workaround, but I have certain habits and I want to stick to them".

It happens occasinally to me as well. This is the spirit that people make find solutions, or do whatever is necessary to reach their goal.

Since I do not have a system with SATA disks, I am afraid I cannot help you any further. Since the day I switched from RH to Debian, I haver never regret this. (Ok, well, I did for 2 or 3 weeks, but then I started to understand what I was doing ). It would be a pity if the Debian community would lose you.

jlinkels

Well... this is how much I want to stick with Debian...

I have removed the SCSI setup from this new machine (which is to be my new server), and have gone with a new IDE boot drive.

This really pisses me off that I have to do this. I can use CentOS and have everything work just fine with SCSI, but I really want to stick with Debian.

So, I am now attempting (and I'm guessing here) my 20th+ install over the last week.

If this works, I'll post back and let you know. If it doesn't... I'm installing CentOS and being done with it.

Is a option to remove the spare disks first.
This way the installer can see only one disk.
Maybe after installing the disks again after installation the drive order rearranges everything again, I don't know.

Maybe fiddle around with the adaptercards in the pci-slots?

Rearrange the boot order in the bios to boot from the specific adaptercard?

You mentioned "software raid".
Did you backup you're data?

I tried to import my software raid disks in a new setup, lost the whole bunch.
Made some big mistakes most likely, anyway if those disks are full, it's very hard to love debian after loosing the data. Is my experience.

Grtz Decep.

You're right, I could do this.

Historically, this would be correct. However, in this situation, it's a limitation of how libata is ordering the drives.

Despite correctly arranging the cards in the BIOS and in the PCI slots... libata coupled with grub still does not allow for IDE/SATA/SCSI to play nicely.

I'm building a new server, then migrating the data from my exsisting server once everything has been set up. Data backup on this new install is irrelevant at this point.

Oh, I'm not trying to import a Windows fakeRAID. I'm setting up a Linux software RAID using the onboard SATA controller.

That sucks losing all of your data. I've had that happen a couple of times...