I have spend lots of time struggling with RAID1. The latest curiosity was that I have been unable to mount a RAID1 filesystem built on 11.1 in 11.2.
Apparently some problem might have been caused by some old partition table. To my recall, the latest attempt involved a BIOS format (which took a couple of days on 1GB drives), clean install of RAID1/ext3 on whole drives using YaST from DVD. I was unable to have new 11.2 system mount the raid.
On Fri, 2009-11-06 at 19:16 +0000, cheongi wrote:
> I have spend lots of time struggling with RAID1. The latest curiosity
> was that I have been unable to mount a RAID1 filesystem built on 11.1 in
> 11.2.
This is NOT an answer. But I have a friend here who’s
messing around with 11.2rc2 and says that is does not
support RAID1 sw boot on his AMD 64bit setup and he says
that it converted ALL seen ext3’s to ext4 (even for OS’s
like Ubuntu installed on the same host in multi-boot) and
he says the grub entries were totally wrong, saying one
thing and pointing to another.
He’s pretty upset… I told him that by definition openSUSE
doesn’t HAVE to work.
On Fri, 2009-11-06 at 19:32 +0000, cjcox wrote:
> On Fri, 2009-11-06 at 19:16 +0000, cheongi wrote:
> > I have spend lots of time struggling with RAID1. The latest curiosity
> > was that I have been unable to mount a RAID1 filesystem built on 11.1 in
> > 11.2.
>
> This is NOT an answer. But I have a friend here who’s
> messing around with 11.2rc2 and says that is does not
> support RAID1 sw boot on his AMD 64bit setup and he says
> that it converted ALL seen ext3’s to ext4 (even for OS’s
> like Ubuntu installed on the same host in multi-boot) and
> he says the grub entries were totally wrong, saying one
> thing and pointing to another.
>
> He’s pretty upset… I told him that by definition openSUSE
> doesn’t HAVE to work.
Also, same person said acpi doesn’t work on his machine
anymore (did on 10.3 and 11.1). That could be something
more serious since newer hardware will require that
acpi works.
Your friend screwed something up. I have installed 11.2 on machine with ext2, ext3, reiserfs & xfs partitions, and NONE of them were converted to ext4. There’s a procedure to convert ext3 to ext4, but my upgrades kept ext3 as ext3, and to get ext4 used I needed to do a clean install into a spare filesystem.
RC versions may very well have issues with software RAID, they will get sorted via updates, for now I should install 11.2 into spare space and try it out, and build RAID sets ater install.
ACPI does generally work, on AMD 64 X2 with Cool N’ Quiet, I have issue with “Suspend to RAM” now, but that is all that broke from 11.1, which as released used to run CPU at only 1 GiHZ due to an erroneously long sampling period > 2 secs for CPU utilisation in the governor.
There’s a huge amount of hardware to support, and a recent kernel like 2.6.31 (only out 6 weeks) is going to throw up a lot of issues once it’s picked up by distro’s, many of them known (and perhaps with solution patches) but which fixes can’t make the deadline for GM.
I have tried raid 1 for /boot in 11.2 final, but with only half success. Creating it is no problem, at boot time grub does it’s job, but then I get a root prompt because fsck can’t check the /boot partition. And that is because /dev/md0 doesn’t exist, and I have to manually issue
that caused the boot process to halt halfway, works like a charm.
(The boot partition is labeled “boot” as you might guess:-))
I have tried to set the boot.md to start in runlevels B, 1,3, and 5, but that didn’t help.
I haven’t tried this before 11.2 final, so I don’t know if it worked before, but unless I’m doing something wrong here, my conclusion is that it doesn’t work in 11.2 final.
I want to create a RAID-5 device for the root partition, but I’m glad I didn’t do that rightaway, because than I wouldn’t have had access to mdadm…
I have wrestling with upgrading 10.3 running on raid 1 (/dev/md0) to 11.2 on a second raid1 (/dev/md2), with some success. 11.2 boots and runs on /dev/md2, but I cannot mount /dev/md0 (and the 10.3 running on /dev/md0 mounts /dev/md2 correctly). So would conclude that 11.2 does work with raid 1.
On OpenSuSe 11.2, I’ve setup md0 and tried to make it the system partition (boot + all system files.) During boot, a boot progress message appears saying that the system is waiting for md0 to appear. This is after the system has already pulled grub’s menu.lst from md0. After awhile it times out and gives up. Have been unable to get past this point.
As an alternate boot, there is a second system image installed on sda2, which boots and after booting can see md0 just fine.
Thinking that the problem might be a lack of a driver in the initrd, I recompiled the kernel with the raid1 driver compiled in (instead of as a loadable module.) This was not sufficient to get past the “waiting for md0” to appear.
In further investigation I discovered that the raid super block has gone through several revisions. As md0 is visible later (after boot) when booted to the system image on sda2, but not during the boot process when booting to md0 – both were installed from same live cd, and had no updates applied, as I couldn’t boot to md0 to install the updates.
It seems likely that the raid driver I compiled in is not the right one for the particular raid super block that the SuSe installer used.
The choices are to find which kernel module actually has to be compiled in, or to try using other versions of the raid super block.
I hope all this helps someone who knows a lot more about raid and SuSe figure out the best approach, meanwhile I’m going to rebuild the kernel with all of the raid and drive mapper drivers compiled in and see if that helps.
> On OpenSuSe 11.2, I’ve setup md0 and tried to make it the system
> partition (boot + all system files.) During boot, a boot progress
> message appears saying that the system is waiting for md0 to appear.
> This is after the system has already pulled grub’s menu.lst from md0.
That file was most probably pulled from one of the two individual
drives, not from the mirror.
> After awhile it times out and gives up. Have been unable to get past
> this point.
The usual problems are - drivers missing, array not being started.
> As an alternate boot, there is a second system image installed on
> sda2, which boots and after booting can see md0 just fine.
That would suggest that md0 does not come up soon enough to be root
filesystem.
> Thinking that the problem might be a lack of a driver in the initrd, I
> recompiled the kernel with the raid1 driver compiled in (instead of as
> a loadable module.) This was not sufficient to get past the “waiting
> for md0” to appear.
So the array is not being auto-started. Can we see the contents
of /etc/mdadm.conf, please?
> It seems likely that the raid driver I compiled in is not the right
> one for the particular raid super block that the SuSe installer used.
> The choices are to find which kernel module actually has to be
> compiled in, or to try using other versions of the raid super block.
Wrt RAID1, the only module you need in the initrd is ‘raid1’.