Is it possible to install + boot int raid 5 device in linux?

remix · l33t Joined: 28 Apr 2004 Posts: 797 Location: hawaii

if not gentoo, then does anyone know if it works with any other distro/kernel?

or does every linux system boot into a real physical device?

https://forums.gentoo.org/viewtopic-t-814245.html
_________________
help the needy

NeddySeagoon · Posted: Fri Feb 05, 2010 8:07 pm Post subject:

remix,

The booting rules are as follows.
1. The BIOS must be able to read the boot loader.
2. The boot loader must be able to read the kernel and initrd, if there is one
3. The kernel (with the aid of the initrd) needs to be able to mount the root filesystem
4. With root mounted you can do anything.

Now to your specific question. There are four sorts of raid5 that linux knows about.
Hardware raid - the raid controller hides the phyisical structure of the raid - it appears as a single device and you use it that way.
Everything just works as the raid card takes care of things.

Fakeraid - the raid is provided partly by the BIOS and partly by the dmraid kernel module. This looks like real hardware raid, hence the name but the kernel can see the underlying devices too. This is the raid you get with onboard SATA controllers.

Windows Dynamic Disk - the kernel supports Windows software raid - I'm not sure if you can boot from it, nor why you would want to

Kernel software raid. Grub1 ignores any raid, which is ok if /boot is not raided or is raid1 but grub fails totally on other kernel raid levels as it does not decode the raid.
I think grub2 understands other raid levels but I've not tested it.

Once you get the kernel loaded, root can be anything. I have root on kernel raid5 and /boot on raid1.
I would have had root inside LVM but that needs an initrd to get LVM up before root can be mounted.

In short, the answer to your question is "Maybe"
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

remix · l33t Joined: 28 Apr 2004 Posts: 797 Location: hawaii

heh, thanks!

well i got 1, and 2 working, failing at 3 where my kernel doesn't know how to mount the root filesystem.

i do have FakeRaid via NVRaid. when i turned it on in bios, livecd (or maybe it's just me) did not know how to see the fake physical device. of course it saw sd[a,b,c,d] but /dev/mapper/ only had control in it. then i fiddled with dmraid, it did "show" me the physical raid named something like nvidia_abiccada, however i was not able to get the device to activate in /dev/mapper/nvidia_abiccada

i also heard from another user that dmraid is not as good as software raid using mdadm.
_________________
help the needy

NeddySeagoon · Posted: Sat Feb 06, 2010 11:01 am Post subject:

remix,

dmraid does work but it has several drawbacks compared to kernel raid.
Its not as robust as kernel raid. If you need to move the raid set to another machine, you need the same hardware and BIOS.
Kernel raid just works.

The only reason to choose dmraid over kernel raid is that Windows must share the raid set.

It sounds like you have a problem in your initrd, not getting dmraid going properly.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

remix · l33t Joined: 28 Apr 2004 Posts: 797 Location: hawaii

i wasn't able to recompile my kernel, kept getting kernel panics.
even when i wanted to emerge a new kernel version, it would crash out on emerge. (python fails)

so i decided to start all over again and reformat my raids. and now everytime i use mdadm to assemble my raids, i get kernel panics, i'm still in livecd mode, my raids are totally empty now so there's nothing to chroot into.

so i create my raids again using mdadm --create and even before i finish formatting, i get kernel panic and livecd system hangs.

i don't know what else i can do now, i feel pretty stuck cause i don't know how to check disk if there is no filesystem to check. it really feels like either the livedvd or my hard drives are broken... isn't that the point of raid5 though? if one fails, it uses the parity of the last to repair? these are brand new disks, brand new everything actually, got each piece new and fresh from newegg.

any other suggestions? i'm going to burn another livedvd now.
_________________
help the needy

remix · l33t Joined: 28 Apr 2004 Posts: 797 Location: hawaii

oh and since my ssh connection died on kernel panic, i can show you only the last messages before it disconnected me

NeddySeagoon · Posted: Wed Feb 10, 2010 6:27 pm Post subject:

remix,

Lets take it easy. Do a standard install on a single drive, if you can.
That may help the diagnostics. If it fails in the same way, try another driver. if that fails too, you have a hardware problem.
Maybe RAM ?
Try memtest from the liveCD. A failure indicates a problem related to the motherboard, not just RAM.
You then need the failure symptoms to narrow down the failure.

If the single drive install works, create your raid sets in degraded mode, with a drive missing.
Copy the single drive install to the degraded raid set. Install grub on the raid, test you can boot the raid.
Lastly, add your single drive install into the raid sets to bring the raids up to strength. This destroys the single drive install.

There - some diagnostics and if it works you yet to practice failed drive replacement too.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

remix · l33t Joined: 28 Apr 2004 Posts: 797 Location: hawaii

Hi Neddy,

I brought in an ATA hd to see if i can just get gentoo to install like normal on one disk, then i wouldn't even mind running everything(boot/root/swap) on that small 160GB and just have /home and /var/www on raid5 disks... failed terribly with Kernel panic - not syncing: VFS: Unable to mount root fs on unknown block(0,0)

very weird cause that's so similar to what i was getting on the non-autodetecting raid setup.

anyway, i found out that the problem wasn't that i didn't enable ext3 fs support in the kernel, cause when i redid the install from scratch again, i found mke2fs -j /dev/sda3 say something about Warning bad superblock or something. and i'm currently running badblocks on that drive, it's riddled with badblocks. so i'm going to decorate my walls with the old ATA harddrive.

now, about this degraded mode... can you go into a bit more detail of how to accomplish this?
step by step, i'm guessing i will reformat (not repartition) the first disk of the 4 disk SATA array.
like normal, ext2 boot, swap, ext3 root, reiserfs home, etc.. and all these partition sizes still mirror the other disks.

so now, i should get the install to work and boot first, correct?

then how do i proceed with copying the drive/partitions to a 'degraded array' (how do i even make a degraded array?) and then add that first drive back in?

i'm very grateful that you have taken the time to help me out with this. this is just beyond my experience and may need a bit of hand holding >.<

thanks!
_________________
help the needy

NeddySeagoon · Posted: Thu Feb 18, 2010 7:20 pm Post subject:

remix,

Try to fix your 160G PATA drive by writing to the entire surface

remix · l33t Joined: 28 Apr 2004 Posts: 797 Location: hawaii

thanks! so i've run across a couple weird things. one, is that my PATA drive is /dev/sda and all my sata drives are b,c,d,e now (instead of a,b,c,d)
i don't think that's a problem, i just think it's odd cause i remember before, PATA drives are always /dev/hd*

and so i started to write to the 160G drive with dd if=/dev/zero of=/dev/hda off the LiveDVD. and i got a general protection fault.

so i think the next step right now would be to get an older x86_84 livecd/dvd cause i know those worked, this one has been giving me kernel panics (off liveDVD environment AND chroot)

then i'll continue with what you wrote step by step.
_________________
help the needy

remix · l33t Joined: 28 Apr 2004 Posts: 797 Location: hawaii

oh, and i'm running Memtest86 right now, it doesn't look too good

http://i49.tinypic.com/2hyjukj.jpg
_________________
help the needy

NeddySeagoon · Posted: Fri Feb 19, 2010 7:25 pm Post subject:

remix,

That has all the hallmarks of a RAM problem as it happens at a fixed addrss.
Memtest errors that cannot be repeated are normally not RAM related. However, more diagnostics are called for.

1/ Remove each stick of RAM and replace in in the same slot. The idea is to 'wipe' the contacts, which often works wonders for about 18 months. If you get the same error at the same address continue with 2.

2/ Rotate Your RAM sticks, i.e. move them all round one slot without bringing any spare slots into play.
Hopefully this will have at least changed the failing addess.

If the failing address has changed - the problem is with a stick of RAM, if not, its something motherboard related.
To isolate a RAM issue, try removing one stick at a time. When memtest runs clean, the not fitted RAM may be faulty.
Test every stick like this. Its not unknown for all combinations of RAM-1 to pass.

If the failing address is unchanged when you rotate the RAM, try abandoning one RAM slot at a time and bringing a spare slot into use, if you have some.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

remix · l33t Joined: 28 Apr 2004 Posts: 797 Location: hawaii

ok so i fixed my RAM problems by reseating it.

and by using an older kernel version, i'm now able to boot off of my raid1 boot partitions and even autodetect the raid arrays, it automatically creates md0,md1,md3,md4,md5 devices (very close to what i had, md1,md2,md3,md4,md5 setup in mdadm.conf so i just adjusted my fstab accordingly)

however, it pauses for a looong time during boot up process, then fails.

here are the errors

before long pause:
http://i48.tinypic.com/2j32np0.jpg

after long pause:
http://i50.tinypic.com/2iuqp82.jpg

any ideas?
_________________
help the needy

NeddySeagoon · Posted: Wed Feb 24, 2010 7:48 pm Post subject:

remix,

Unable to open initial console is a bad sign on its own. I can think of three causes.
There is no console configured in the kernel. You have a choice of two sorts but its not an error to leave them both out.
The system still works.

You have a udev that is too old/new for your kernel and its not making /dev/console

You have an older udev that requires a static /dev/console and its missing

For the panic
At the the grub splash screen press e.
Follow the on screen instructions to edit the kernel line.
Append options like noapic noapci both separately and together.

Please post your kernel .config somewhere. You may email it to me and I'll post it for you.
Its the file at /usr/src/linux/.config
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

remix · l33t Joined: 28 Apr 2004 Posts: 797 Location: hawaii

here is my kernel config. http://pastie.org/841219

what does noapic and noapci do?

and i should boot up with them with a kernel that has a console configured, yes?
_________________
help the needy

NeddySeagoon · Posted: Wed Feb 24, 2010 9:58 pm Post subject:

remix,

One turns off the Advanced Peripheral Interrupt Controller, the other turns off the Advanced Power Control Interface, which are good thinks to have working but can cause issues if the kernel is not set up right.

I'll look at your kernel tomorow
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

NeddySeagoon · Posted: Thu Feb 25, 2010 8:31 pm Post subject:

remix,

There is a lot of junk in your kernel but I don't see anything that should prevent it booting.

You have two drivers built the your PATA interface, the old IDE one and the new libata one, so its unlikely that PATA will work at all.
There are a few of my kernel config files here, all for gentoo-sources.

My lspci looks like