Guest FireStorm9 Posted October 12, 2004 Report Share Posted October 12, 2004 Greetings! I'm having grief trying to install MD10.1 from scratch on a P4 2.6 Abit IS7 with ATI video card and an additional 3com NIC as a file/server for the house. I've been running linux since RedHat 7, (so not that long) but I've repeatedly used the same setup: 2 ide hard drives, partitioned and mirrored, and have had pretty good luck with that setup until now. MDK10.1 will not boot. I get mdadm unloading, then unbinding hda1 and hdd1, then the kernel panics because /dev/md1 (root) is not there. The partitions are simple: 2 Maxtor 160gb drives on the standard ide channels: Note: for all matters, hdd is the other hard disk and it is exactly the same partitioning schema as hda. hda1: 1024mb linux raid autodetect hda2: 4096mb swap hda5: extended partition hda6: 30720mb linux raid autodetect hda7: rest of disk linux raid autodetect These map to: hda1 & hdd1: /dev/md0 (/boot) hda6 & hdd6: /dev/md1 (/) hda7 & hdd7: /dev/md2 (/home) During the installation, I choose custom partitioning, set up both disks to have identical partitions, then click the "add to raid" button, select raid 1, type ext3, and assign the mount points. The disk configuration utility formats the disk, and life is good; the installation continues. When it reboots, I get a panic error before the kernel even loads, so I don't even get any logging. I've replicated this problem on THREE DIFFERENT MACHINES, including a VM Ware machine (just for grins). I must be totally screwing up here, does anyone have any advice? Many thanks in advance, -=FireStorm Quote Link to comment Share on other sites More sharing options...
devries Posted October 12, 2004 Report Share Posted October 12, 2004 Im not really an expert but when you say you get a panic before the kernel loads does that mean a problem with the bootloader? If so check the /etc/lilo.conf. Quote Link to comment Share on other sites More sharing options...
Guest FireStorm9 Posted October 12, 2004 Report Share Posted October 12, 2004 (edited) The only output I can see (using framebuffer) is: Before you note the hdb/hdd disparity, the hdd is on the home box, the hdb is on the vmware machine. Both behave identically. ==== Please ignore any misspellings or such. I had to hand-type this stuff in ==== -------begin------ md :do_md_run() returned -22 md: md0 stopped. md: unbind<hdb1> md: export_rdev(hdb1) md: unbind<hda1> md: export_rdev(hda1) md: ... autorun DONE. RAMDISK: Compressed image found at block 0 VFS: Mounted root (ext2 filesystem). Red Hat nash version 4.1.9mdk starting Loading raid1.ko module md: raid1 personality registered as nr 3 Loading jbd.ko module Loading ext3.ko module Mounting /proc filesystem Mounting sysfs Creating device files mounting tmpfs on /dev starting udev Activating md devices md: md1 stopped mdadm: no devices found for /dev/md1 Creating root device Mounting root filesystem EXT3-fs: unable to read superblock mount: error 22 mounting ext3 flags defaults well, retrying without the option flags EXT3-fs: unable to read superblock mount: error 22 mounting ext3 well, retrying read-only without any flag EXT3-fs: unable to read superblock mount: error 22 mounting ext3 pivotroot: pivot_root(/sysroot, /sysroot/initrd) failed: 2 umount /initrd/sys failed: 2 umount /initrd/proc failed: 2 Initrd finished Freeing unused kernel memory: 216k freed Kernel panic: No init found. Try passing init= option to kernel. -------end------ Contents of /etc/lilo.conf -------begin------ # File generated by DrakX/drakboot # WARNING: do not forget to run lilo after modifying this file default="linux" boot=/dev/hda1 map=/boot/map keytable=/boot/us.klt prompt nowarn timeout=100 message=/boot/message menu-scheme=wb:bw:wb:bw image=/boot/vmlinuz label="linux" root=/dev/md1 initrd=/boot/initrd.img append="devfs=nomount acpi=ht resume=/dev/hda5 splash=silent" vga=791 read-only image=/boot/vmlinuz label="linux-nonfb" root=/dev/md1 initrd=/boot/initrd.img append="devfs=nomount acpi=ht resume=/dev/hda5" read-only image=/boot/vmlinuz label="failsafe" root=/dev/md1 initrd=/boot/initrd.img append="failsafe acpi=ht resume=/dev/hda5 devfs=nomount" read-only -------end------ Thanks again, -=FireStorm Edited October 12, 2004 by FireStorm9 Quote Link to comment Share on other sites More sharing options...
Guest FireStorm9 Posted October 13, 2004 Report Share Posted October 13, 2004 So NOBODY on the planet has replicated this? Or identified it as a bug? Is it possible that NO ONE on the planet has tried to install 10.1 using software raid1? Is that a horrible configuration? Quote Link to comment Share on other sites More sharing options...
Guest fraynet Posted October 21, 2004 Report Share Posted October 21, 2004 From what I know you can't use software raid on your /Boot partition. I had the same problem when installing Raid 1 with Suse 9.1 I read in a Software Raid howto that mentioned /Boot can't be raided. If anything ever did go wrong it wouldn't be that difficult to fix the /boot anyways. Hope this helps -Tim Quote Link to comment Share on other sites More sharing options...
Guest FireStorm9 Posted October 21, 2004 Report Share Posted October 21, 2004 Tim, Thanks for your reply. I respectfully disagree though, as with a Fedora 2 installation, the partitions are md0, 1, and 2, and work beautifully, the first time out. Similarly, on a Mandrake 9.2 machine, the partitions are this: /dev/md/0 on /boot type ext3 (rw) /dev/md/1 on / type ext3 (rw) /dev/md/2 on /home type ext3 (rw) And that works fine also. I contend that there's something completely porked with the scripts that run after installation, that orders things wrong, or raid1 doesn't get put into the kernel at boot time......I don't know, I just gave up and went to Fedora. It's a shame really, because Mandrake seems much faster to me and I'm quite used to where Mandrake puts things. :( Thanks again! Quote Link to comment Share on other sites More sharing options...
Guest maroelawerner Posted October 26, 2004 Report Share Posted October 26, 2004 (edited) So NOBODY on the planet has replicated this? Or identified it as a bug? Is it possible that NO ONE on the planet has tried to install 10.1 using software raid1? Is that a horrible configuration? <{POST_SNAPBACK}> Hi, I got the same problem that you have. The only difference is that I have four RAID 0 partitions, but the output is the same except for were your one line reads "loading raid1.ko module" , mine read raid0.ko module. Ralph Edited October 26, 2004 by maroelawerner Quote Link to comment Share on other sites More sharing options...
Guest anon Posted October 26, 2004 Report Share Posted October 26, 2004 So NOBODY on the planet has replicated this? Or identified it as a bug?Is it possible that NO ONE on the planet has tried to install 10.1 using software raid1? Is that a horrible configuration? <{POST_SNAPBACK}> This info any help? RAIDA new tool, dmraid (Device-Mapper Raid tool) was included in the distribution. It discovers, activates and displays properties of software RAID sets (ie. ATARAID) and contained DOS partitions using the device-mapper runtime of the 2.6 kernel. The following ATARAID types are supported on Linux 2.6: * Highpoint HPT37X * Highpoint HPT45X * Intel Software RAID * LSI Logic MegaRAID? (basic support) * Promise FastTrack? * Silicon Image Medley Quote Link to comment Share on other sites More sharing options...
Guest konradj Posted November 2, 2004 Report Share Posted November 2, 2004 I have the same problem, with 10.1. Did a cline install width software raid. 2 Maxtor 160gb sata /dev/md0/ root file system and boot /dev/md1/ swap /dev/md2/ data1 /dev/md3/ data2 My system boots, and md0 works fine. but md1-md3 are not created in /dev, and that makes the system unable start md1-md3. I think the problem is some where in udev, sins /dev/md1-3 are not created. If you put root and /boot on /dev/md0, you probably will be able to boot your system. Anyone haw some experience? How can I manually force udev make /dev/md1-3? Quote Link to comment Share on other sites More sharing options...
Guest sonnenschein987 Posted November 12, 2004 Report Share Posted November 12, 2004 Hi, I have similar problems with the software raid, and I am close to moving to a hardware raid, but I have some ideas. Mandrake 10.x comes with mdadm as standard. You can construct the raids, but after a reboot they fail, because fsck does not work. Previous I used raidtools and my raids worked fine, o.k. if they worked perfectly I wouldn't have to reinstall my computer after a major crash, but as a matter of fact now I have 10.1 running and it's quite better than 10.0. And coming to another question, mandrake and booting from raid. I am not sure if that will work, because the raid system is started after linux is initialized, so the init can't find the raid. Maybe there are some kernels where the raid is initialized very early, than it can work. But to come to a conclusion. I would like not to use raids, but even if not official verified, my machine is much faster when running in raid 1 (not / or /boot) but file partitions. sorry, but no other idea, but if someone finds a reliable howto for mandrake 10 and mdadm it would be nice to hear about it. Thomas Quote Link to comment Share on other sites More sharing options...
Guest FireStorm9 Posted November 12, 2004 Report Share Posted November 12, 2004 I've messed and messed and messed with this. When booting off the rescue disk, I can do the following: modprobe raid1 mknod /dev/md0 b 9 0 mknod /dev/md1 b 9 1 mknod /dev/md2 b 9 2 mount /dev/hda /mnt cp /mnt/etc/mdadm.conf /tmp umount /mnt mdadm --assemble -c /tmp/mdadm.conf /dev/md0 mdadm --assemble -c /tmp/mdadm.conf /dev/md1 mdadm --assemble -c /tmp/mdadm.conf /dev/md2 At this point, all raid volumns are up and running normally, so I know two things: The volumes are correct, and mdadm works as it should. What SEEMS to be happening is either: 1) raid1 is not compiled into the kernel, or 2) raid1 gets loaded AFTER udev tries to create the /dev/md? devices. I've tried futzing around with mkinitrd and crap like that but I'm just not proficient enough to make it work. No matter what I do I ALWAYS get a kernel panic. I have successfully got Mandrake 10.0 to work, but it's using devfsd instead of udev. It just HAS to be a UDEV problem in the order it's loading things. The real smoking gun is that md autodetects the arrays at boot time and tries to start them, but get's a -22 error with "Personality 3 not loaded". Then later in the boot process, you see that it's loaded just before udev. <rant> I've had to move to Fedora 3 (which I HATE) as a result of this complete broken boot process. I'd LOVE for Mandrake to figure this out, but I have been unsuccessful at getting this to work, or to find anyone that knows enough to help me make it work. </rant> A possible helpful thing would be to get DETAILED debugging messages as the kernel loads, but I don't know of any append= switch to accommodate that. Any experts out ther have any insight? Thanks! Quote Link to comment Share on other sites More sharing options...
Barge Posted November 21, 2004 Report Share Posted November 21, 2004 I have found the exact same problem with 10.0 when trying to do this as well. My natural preference for servers I was putting in at work is mirrorred RAID 1 on the boot disk(s). I am presuming the same problem has persisted to 10.1 (I haven't tried it yet). When I was doing work on this I found when reverting to my 9.2 CD's that it worked fine. I did a lot of messing around and reinstalls to see how this problem was happening: - installing 9.2 to RAID1 partitions works fine - dropping in a 10.0 DVD and using the "upgrade" function following a successful 9.2 install (which I generally avoid but used for testing this) worked and left me with a fully functional 10.0 install on RAID1 partitions. After I tried that I was surprised how the 9.2 worked but the 10 didn't... not something I have generally run into with Mandrake installs. You can try an upgrade from 9.2 or jsut creating the filesystems with 9.2 and rebooting to boot a version 10.1 install- maybe see how these procedures go. Also watch carefully when you create your RAID1 filesystems- I noticed that one time when creating them, the version 10 install did not correctly set the chunk size on one of the partitions- one of them did not have a chunk size specified, and when the installer tried to mkraid in the bankground it consequently failed. Quote Link to comment Share on other sites More sharing options...
Guest nicolas Posted November 22, 2004 Report Share Posted November 22, 2004 I'have the same problem with Mandrake 10.1. I solved the problem ricompiling the kernel with all the software RAID modules built in (NO MODULE FOR SOFTWARE RAID). Afther reboot the new kernel work fine. Quote Link to comment Share on other sites More sharing options...
Guest privacy Posted December 23, 2004 Report Share Posted December 23, 2004 I'have the same problem with Mandrake 10.1. I solved the problem ricompiling the kernel with all the software RAID modules built in (NO MODULE FOR SOFTWARE RAID). Afther reboot the new kernel work fine. <{POST_SNAPBACK}> I have exactly the same problem as above described. When i use the rescue CD it mounts /dev/hda2, one disk of /dev/md1, on /mnt. But there is no raidtab file shown in /mnt/etc/. Is this also a problem or could the file be located somewhere else? cheers chris Quote Link to comment Share on other sites More sharing options...
Guest nicolas Posted December 23, 2004 Report Share Posted December 23, 2004 I have installed Mandrake 10.1 with the normal installation process (using graphic installation). When I make partitions (diskdrake) I made one big raid 1 (mirroring) partition for root filesystem using the appropriate installer function (look at diskdrake for detail, it's just point and click). I have made two identical swap partition on the two hard drive. Diskdrake produce a warning about the difficult to start the system from a raid partition but you can ignore it. Continue with setup. When you reboot the system the kernel cant mount the root filesystem because the raid kernel module in initrd have problems to load correctly. But if you can start the kernel (see the message of FireStorm9 post Nov 12 2004, 06:52 PM in this thread) the raid work fine. All other initialization files are build correctly by the installer, and no modify are need. I have solved in this mode: boot with rescue disk and mount the raid partition (see the message of FireStorm9 post Nov 12 2004, 06:52 PM in this thread). Install the kernel source from media. Copy the config file from /boot to the kernel source directory (.config). Run the menuconfig utility to change the kernel configuration. AT THIS POINT SEARCH FOR ALL THE SOFTWARE RAID MODULE and make them no modules, but fixed compiled in kernel. SEARCH AGAIN for the module of your root filesystem (raiserfs.ko in my installation) and make it fixed too. Rebuild and install new kernel as usual. Now you can reboot the system. When the kernel boot the autodetect of raid volume work fine and the root raid volume will be mounted. The problem I have detected come from a non correct load of modules in initrd image (paraphs i have tried to rebuild initrd but without success); the only chance is to avoid the need of loading external module during the startup. Excuse for my bud english, I'm italian. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.