Enabling RAID autodetection in kernel

km · n00b Joined: 17 Jul 2002 Posts: 10

Hello,

I am trying to setup software RAID on a brand new Gentoo 1.2 install - following the software RAID installation instructions. Everything works fine, except that I can't seem to find where to enable the RAID autodetection option when configuring the kernel. I thought it was supposed to be in the RAID/LVM section, but I don't see it there. So, of course, my system is not able to mount the root filesystem - GRUB comes up, but when I hit enter, I get a kernel panic error.

Am I missing something?

Thanks!

BillyD · Guru Joined: 05 May 2002 Posts: 323 Location: Australia

Under ATA/IDE/MFM/RLL support, you can select options for promise and higpoint controllers. I have a promise fasttrack controller, and the options I have selected to enable support for it are:
PROMISE PDC202{46|62|65|67|68|69|70} support
-- Special Fasttrack Feature

and

Support for IDE Raid controllers
Support Promise software RAID (Fasttrack(tm))
_________________
We used to have hominid cousins that were vegetarian. The palæontological record suggests that our ancestors killed them and ate them.

km · n00b Joined: 17 Jul 2002 Posts: 10

I'm not at that computer right now, but I don't think that's what I need - I'm using SCSI drives (with an Adaptec controller), but it's not a RAID controller. I'm using software RAID.

... unless maybe I didn't understand your reply...

If you look here http://www-106.ibm.com/developerworks/linux/library/l-raid1/?dwzone=linux there's an article (by Daniel Robbins, actually) that has a screen shot of what I'm talking about. Under the Mulit-device support section, there is an option for Auto Detect support - which I don't have when I do make menuconfig (actually, I don't have the Boot support option either).

Does anyone know why? I already tried a vanilla kernel, but I got the same thing.

Help!

chatwood2 · Posted: Thu Jul 18, 2002 2:09 pm Post subject:

I just checked one of my gentoo machines with a RAID in it, and that option "Auto Detect Support" isnt there either. Under "Mutiple devices driver support (RAID and LVM)" I only have "RAID support", "RAID-0", and "RAID-1" compiled into the kernel, no modules.

Going back to your problem. While in fdisk did you change all partitions that are going to be in the RAID into type "fd" (the label under "System" when printing the partiton table should be "Linux raid autodetect.")

What does your /boot/grub/menu.1st look like? Specifically what device does root point to?

- Chris

km · n00b Joined: 17 Jul 2002 Posts: 10

I have the same kernel settings as you: RAID Support, RAID0 & RAID1 compiled in - no modules.

In fdisk, I set all of the partitions to fd (and I double checked them to make sure I didn't forget!) - they all say Linux raid autodetect. Actually, I used the same partitioning scheme as in your docs (except that my drives are SCSI):

chatwood2 · Posted: Thu Jul 18, 2002 6:24 pm Post subject:

That -force on the mkraid worries me, you shouldn't have had to do that. But I guess if it did start and sync ok I guess it doesn't matter. During the install did you (after making the RAIDs)

rommel · Posted: Thu Jul 18, 2002 7:24 pm Post subject:

auto-detection is actually created when you set the type of partition ...you should ahve created the partitions you wanted to use as part of your raid setup the changed the partition type to fd...this is lsr autodetection...then in the kernel just enable the raid level you are you using...you can enable all of them but if your only using raid 0 why have linear and mirroring compiled in.

Guest · Posted: Fri Jul 19, 2002 2:41 am Post subject:

OK, just to make sure I didn't screw anything up, I started from scratch again. I ran fdisk on both drives, and just to be sure, I removed all the partitions wrote the (blank) partition table, then started fdisk again and re-built the partition tables, and verified that they were all set to type fd. I checked the RAID status (/proc/mdstat) after each mkraid command, and everything looked good. Here is what /proc/mdstat looked like after all the disks had finished syncing:

km · n00b Joined: 17 Jul 2002 Posts: 10

Oops. The previous (guest) reply was from me - I guess I wasn't logged in...

Anyway, one other thing: is it possible that the drives and SCSI card I'm using are too old to support the autodetection stuff? They're good (adaptec card, seagate drives), but they're pretty old... I don't think it should make a difference, but I just thought I'd ask.

rommel · Posted: Fri Jul 19, 2002 4:12 am Post subject:

i am not sure what your doing wrong ... i have set up raid 0 several times on both ide and scsi drives...right now i am running gentoo on three scsi seagate cheetahs....i partitioned the drives with three 500mb primary partitions on each drive sda1 is boot sdb1 and sdc1 are swap...then i created extended partitions with the remainder of each drive and one logical partiton in each extended partition so that /dev/md0 consists of sda5,sdb5 and sdc5

/etc/raidtab looked like this

raiddev /dev/md0
raid-level 0
nr-raid-disks 3
persistent-superblock 1
chunk-size 32
device /dev/sda5
raid-disk 0
device /dev/sdb5
raid-disk 1
device /dev/sdc5
raid-disk 2

then when you issue the mkraid it reads from this file...but i know your getting this far...as for the force option having to be executed it seems that the superblock adn raid array is still being seen thats why it is asking you to force it...you should try 'mkraid --really-force /dev/md?...this will ensure that you are over writting the old raid if its persisting for some reason.

you could also try erasing the /dev/md? partition by using if=/dev/zero of=/dev/md? bs=1k , i think thats the right code

if you feel like it post your grub config the raidtab your creating and fstab.

km · n00b Joined: 17 Jul 2002 Posts: 10

When I say I'm using the -force option to mkraid, I'm actually using --really-force, so that's covered. I haven't tried explicitly erasing /dev/md? - maybe I'll try that next. If you look a couple of posts back, you'll see my Grub configuration. Here's my raidtab:

chatwood2 · Posted: Fri Jul 19, 2002 5:37 pm Post subject:

I don't think the problem is with your RAID, I think it is with your filesystem. You shouldn't have had to do the -force option when making the RAIDs, what was on these drives before this gentoo install? You might be getting XFS and ext2 errors because of rements of past filesystems. Have you tried zero'ing the drives?

Also, what kernel options have you enabled under "file systems"?

And since we are dealing with SCSI here, your sure that you chain is terminated and that all cables are in good condition? What SCSI speed are you using (wide, Ultra Wide, Ultra2, etc)?

Sorry about not having any answers for you, but we are elimating possible problem areas.

- Chris

rommel · Posted: Fri Jul 19, 2002 5:50 pm Post subject:

chatwood makes a couple of good suggestions....the zero'ing out of the partitions would be a good idea and obviously checking your hardware would be a must.

i really cant see anything that shouldnt work for you from what you have posted....although i have not tried to mirror the boot partition myself ,i know it can be done.

also when you have issued the --really-force command did you first attempt to just use raidstart /dev/md? ...... that should have worked because the reason to use the -foce option is to over-write an existing raid device

km · n00b Joined: 17 Jul 2002 Posts: 10

I'm gonna check what filesystems I have compiled into the kernel when I get home, and see if there's any extraneous stuff in there. I already checked, re-checked, and checked again all of the SCSI stuff (I've been burned by SCSI before) and everything looks good. I agree with the -force thing, I'm gonna wipe those drives when I get home and start over (again!) and see what I get. I've used those drives for Linux before (never for RAID, though) so I can't be sure what was on there when I started - although I'm sure that EXT2 was on there at some point, but definitely never XFS. Hmmm...

When I -force'ed mkraid, I didn't try raidstart first (at least not the very first time), but I do remember the error I got from mkraid when I tried it without -force said something about already finding EXT2 on the drive. I didn't think much about it at the time - I figured mkraid would just write over whatever was there, but that may be part of the problem: if I just finished telling fdisk to make all the partitions type fd, why would mkraid see EXT2 anywhere on the drive?

Any suggestions on the best way to zero the drives? I was thinking about doing a low-level format through the SCSI utilities. That ought to do it, huh?

As usual, thanks for the help and suggestions!

rommel · Posted: Sat Jul 20, 2002 1:51 am Post subject:

i think it sees the persistent superblock and assumes that there is or was a raided device and warns you.

but if you write the zero mask to the entire drive and then repartition and set the raidtab again you should be ok....create boot as a primary and swap as a primary then / as a logical drive inside an extended partition...i am assuming that boot and swap are relatively small...mine are

Guest · Posted: Sat Jul 20, 2002 3:57 am Post subject: ISO file system ???

Comparing with common mount errors at command line, I believe kernel tries all possibilities
when autodetected. Look at error reports when fstab is set to default mount
floppy or cdrom drives and they are not set correctly or MT --It tries all f/system for drive type, enabled in kernel - before giving up - after several passes.
The odd >>ext2, xfs<< may be same type of attempts. Note order 3 - 2- iso-xfs
ever had any hard drive using cdrom iso9660 specs ?
Zero - isn't that same as dd useage for zeroing a file directory for swap useage
Not swap (partition) If so, it would be dd if=/dev/zero of=/ (target)
Syntax aside, why would it be required, one drive- doubtful, all -??? (grasping 4 straws)
After install - before reboot - have U tried an unmounted e2fsck, ( &/or w/8193)
I would suspect loader has been pointed at wrong path for getting file info ??
Think of alternate ways to check, re re re the same things may relieve stress
but only reconfirms those things still don't cure anything.
If really frustrated, do something else - totally non-computer related. The
sub-concious often works best on its own.
OO curiosity no b/up, extra levels of errors being introduced why Software raid ?? Good hunting 3x9

Guest · Posted: Sat Jul 20, 2002 4:17 am Post subject: persistent superblock

Re persistent theory
reposition hard drives to eliminate hitting same bad block ???
Quick install of any other O/sys (un-Gen2) even old [ A:\ > sys. com C: \ ]
to transfer ios.sys & msdos.sys from Dos rescue floppy (two minutes) would rule out anything SCSI related one drive only needed ?

3x9

Guest · Posted: Sat Jul 20, 2002 4:33 am Post subject: autodetect

Damn this login & out thing, every time go to other site, lose page for forum !

Anyhow, if not detected properly, how would you be able to open drives for
partitioning, then mke2fs ?

km · n00b Joined: 17 Jul 2002 Posts: 10

OK, I low level formatted both drives, then started over, and I got the same thing. I'm gonna try installing another OS on the drives and make sure they boot OK, as soon as I get a chance.

Is there any reason you suggested that I make the boot and swap partitions primary and the rest extended? I've been making all 3 partitions primary each time - do you think that may be part of the problem? BTW I have my other Linux box setup with 3 primary partitions, and it works fine.

Thanks

chatwood2 · Posted: Mon Jul 22, 2002 1:33 pm Post subject:

I run all primary partitions on my machines.

Last night I installed gentoo on a machine following my software RAID how-to and ran into zero problems. So, its not like there is a huge problem with the docs, there is just some strange issue with your machine.

- Chris

km · n00b Joined: 17 Jul 2002 Posts: 10

OK, thanks. I thought all primary was OK.

Yeah, I have absolutely no problems with your docs either - I agree that it's definitely something screwy with the setup I'm running. As much as I hate to say it, I'm afraid it might be hardware at this point. I'm gonna see if I can scrape up the stuff I need to try it again with different drives, and/or a different SCSI controller - or maybe I'll try it on another machine, maybe it's something weird with the BIOS...

Yama · Posted: Sat Jul 27, 2002 4:34 am Post subject: You are not alone

I don't think the problem is hardware. I have a brand new machine that seems to work fine with Windows and with Gentoo without software RAID. When I try to use software RAID, however, things get messed up, and I get the same problem as you. My drives are ext2, ReiserFS and XFS, so I'm using an XFS kernel (can the Gentoo kernel handle XFS?). I see no kernel option for enabling RAID autodetection, so I get a "cannot mount root FS" error at boot. Here's my /etc/raidtab:

Yama · Posted: Sat Jul 27, 2002 4:39 am Post subject: Re: You are not alone

Yama · Posted: Sat Jul 27, 2002 2:23 pm Post subject: All fixed (I think)

I think everything's under control now. I recently discovered that XFS doesn't play nicely with Gentoo's kernel optimisations, so I decided to reformat my XFS filesystems to ReiserFS (after first using dd to wipe them clean). I then emerged the Gentoo kernel sources and configured my own kernel. There is no option for RAID autodetection, but it appears that it is bundled in if you compile RAID directly into the kernel (not as a module). My RAID devices worked perfectly after installing the new kernel and rebooting.

chatwood2, thanks for the excellent tutorial!

rommel · Posted: Sat Jul 27, 2002 5:46 pm Post subject:

i run the vanilla kernel and never have trouble...everytime i would run a gentoo or jp kernel something would get messed up...so to get one added thing i would be loosing two it seemed.

xfs was a miserable performer and reisersfs blew up on me and i lost the ability to boot, so this last install i went to vanilla kernel and ext3 and everything is stable as well as fast.

on my lsr raid 0 array running hdparm i get

rommel # hdparm -tT /dev/md0

/dev/md0:
Timing buffer-cache reads: 128 MB in 0.47 seconds =272.34 MB/sec
Timing buffered disk reads: 64 MB in 0.82 seconds = 78.05 MB/sec