Re: mdadm error - superfluous RAID member
On Wed, Jun 13, 2012 at 6:36 PM, Steve Dowe s...@warpuniversal.co.uk wrote: On 13/06/12 23:15, Tom H wrote: Since metadata 1.1 or 1.2 stores the metadata at the beginning rather than at the end, perhaps using a partitioned mdraid device with that metada works with squeeze. Good idea. I'll boot it up with a live CD and report back soon. I don't think that you can change metadata version without a re-format. -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/CAOdo=SzjZokQ6xiObx9W_X-uvOkipZ=UGWpihtYW41x=ym-...@mail.gmail.com
mdadm error - superfluous RAID member
Hi, I'm trying to re-use an older server, installing squeeze (6.0.5). I'm using software RAID and LVM on the machine (details below). But I must be doing something wrong with the disk set up stage in the installer, as when it boots I see an error flash up quickly: error: superfluous RAID member (5 found) It appears that the initramfs then gets loaded, the RAID detection fails and it then looks for the LVM volume group, which it can't find (as the LVM group exists on the RAID device). I see this output: Loading, please wait... mdadm: No devices listed in conf file were found. Volume group vgbiff not found Skipping volume group vgbiff Unable to find LVM volume vgbiff/lvroot same messages appear but for lvswap Gave up waiting for root device snip ... It then drops me into the BusyBox shell, with initramfs prompt. I can then activate the RAID simply by doing (initramfs) mdadm --assemble --scan mdadm: /dev/md/0 has been started with 5 drives and 1 spare. and then activate the volume group, using: (initramfs) vgchange -a y 2 logical volume(s) in volume group vgbiff now active Exiting the busybox shell then boots the system. The basic configuration is: - Xeon (64-bit capable) w/4GB RAM - PCI SCSI controller - 6 x 73GB SCSI drives During install, on each drive I created a 500MB primary partition (with /dev/sda1 being for /boot) and then a second partition for Linux s/w RAID (label set to fd). In /dev/md0 I then created a LVM partition, and set up the volume group to contain two volumes - one for swap, and one for /. /dev/md0 is comprised of 5 drives running in RAID5, with one hot spare. During installation, I took pains to wipe all the drives and create all partitions anew. When booted, I checked /etc/default/mdadm. The values INITRDSTART='all' and AUTOSTART=true are both set. I also set VERBOSE=true to give me more output when creating a new initramfs. I checked the contents of /etc/mdadm/mdadm.conf - which seems fine. I then issued update-initramfs -vu, and saw the following: I: mdadm: using configuration file: /etc/mdadm/mdadm.conf I: mdadm: will start all available MD arrays from the initial ramdisk. I: mdadm: use `dpkg-reconfigure --priority=low mdadm` to change this. and the last output before cpio builds the initial ramdisk is Calling hook dmsetup - so, in my limited knowledge, this suggests the drive mapper is incorporated into the initramfs also. When I take a peek into /boot/grub/grub.cfg I see: insmod raid insmod raid5rec insmod mdraid insmod lvm in the 00_header section. I'm running low on ideas now. Re-installing grub doesn't help. Running update-grub simply dumps out many more of those error messages: error: superfluous RAID member (5 found). repeats 17 times So it does point to grub being at fault somewhere, rather than the initrd. Have I missed something blindingly obvious? Thanks again, Steve -- Steve Dowe Warp Universal Limited http://warp2.me/sd -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/4fd8539b.9080...@warpuniversal.co.uk
Re: mdadm error - superfluous RAID member
On Wed, Jun 13, 2012 at 3:47 AM, Steve Dowe s...@warpuniversal.co.uk wrote: Hi, I'm trying to re-use an older server, installing squeeze (6.0.5). I'm using software RAID and LVM on the machine (details below). But I must be doing something wrong with the disk set up stage in the installer, as when it boots I see an error flash up quickly: error: superfluous RAID member (5 found) It appears that the initramfs then gets loaded, the RAID detection fails and it then looks for the LVM volume group, which it can't find (as the LVM group exists on the RAID device). I don't believe you can boot from a striped volume (raid5 being a stripe + parity). I found some instructions that may allow this to work but requires packing a non-standard initrd: http://nil-techno.blogspot.com/2009/02/booting-fakeraid-raid5-linux-half-assed.html -- Chris -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/CAOEVnYuG3eipFjkxRROp=gRkYr8NhFuJ6K=xv9=izayogyr...@mail.gmail.com
Re: mdadm error - superfluous RAID member
On Wed, Jun 13, 2012 at 4:47 AM, Steve Dowe s...@warpuniversal.co.uk wrote: I'm trying to re-use an older server, installing squeeze (6.0.5). I'm using software RAID and LVM on the machine (details below). But I must be doing something wrong with the disk set up stage in the installer, as when it boots I see an error flash up quickly: error: superfluous RAID member (5 found) It appears that the initramfs then gets loaded, the RAID detection fails and it then looks for the LVM volume group, which it can't find (as the LVM group exists on the RAID device). I see this output: Loading, please wait... mdadm: No devices listed in conf file were found. Volume group vgbiff not found Skipping volume group vgbiff Unable to find LVM volume vgbiff/lvroot same messages appear but for lvswap Gave up waiting for root device snip ... It then drops me into the BusyBox shell, with initramfs prompt. I can then activate the RAID simply by doing (initramfs) mdadm --assemble --scan mdadm: /dev/md/0 has been started with 5 drives and 1 spare. and then activate the volume group, using: (initramfs) vgchange -a y 2 logical volume(s) in volume group vgbiff now active Exiting the busybox shell then boots the system. The basic configuration is: - Xeon (64-bit capable) w/4GB RAM - PCI SCSI controller - 6 x 73GB SCSI drives During install, on each drive I created a 500MB primary partition (with /dev/sda1 being for /boot) and then a second partition for Linux s/w RAID (label set to fd). In /dev/md0 I then created a LVM partition, and set up the volume group to contain two volumes - one for swap, and one for /. /dev/md0 is comprised of 5 drives running in RAID5, with one hot spare. During installation, I took pains to wipe all the drives and create all partitions anew. When booted, I checked /etc/default/mdadm. The values INITRDSTART='all' and AUTOSTART=true are both set. I also set VERBOSE=true to give me more output when creating a new initramfs. I checked the contents of /etc/mdadm/mdadm.conf - which seems fine. I then issued update-initramfs -vu, and saw the following: I: mdadm: using configuration file: /etc/mdadm/mdadm.conf I: mdadm: will start all available MD arrays from the initial ramdisk. I: mdadm: use `dpkg-reconfigure --priority=low mdadm` to change this. and the last output before cpio builds the initial ramdisk is Calling hook dmsetup - so, in my limited knowledge, this suggests the drive mapper is incorporated into the initramfs also. When I take a peek into /boot/grub/grub.cfg I see: insmod raid insmod raid5rec insmod mdraid insmod lvm in the 00_header section. I'm running low on ideas now. Re-installing grub doesn't help. Running update-grub simply dumps out many more of those error messages: error: superfluous RAID member (5 found). repeats 17 times So it does point to grub being at fault somewhere, rather than the initrd. Maybe this bug: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=610184 (What mdraid metadata version are you using?) -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/CAOdo=Sx=gbw25x8__n6q5fob07fy-p6_5r6onbv8tgwdszv...@mail.gmail.com
Re: mdadm error - superfluous RAID member
On Wed, Jun 13, 2012 at 12:40 PM, Christofer C. Bell christofer.c.b...@gmail.com wrote: On Wed, Jun 13, 2012 at 3:47 AM, Steve Dowe s...@warpuniversal.co.uk wrote: I'm trying to re-use an older server, installing squeeze (6.0.5). I'm using software RAID and LVM on the machine (details below). But I must be doing something wrong with the disk set up stage in the installer, as when it boots I see an error flash up quickly: error: superfluous RAID member (5 found) It appears that the initramfs then gets loaded, the RAID detection fails and it then looks for the LVM volume group, which it can't find (as the LVM group exists on the RAID device). I don't believe you can boot from a striped volume (raid5 being a stripe + parity). I found some instructions that may allow this to work but requires packing a non-standard initrd: http://nil-techno.blogspot.com/2009/02/booting-fakeraid-raid5-linux-half-assed.html grub2 can handle /boot on mdraid raid5 (and possibly dmraid raid5 too). From a raid5 VM: [root]# grep md0 /proc/mdstat md0 : active raid5 sdc1[2] sdb1[1] sda1[0] [root]# [root]# [root]# mount | egrep -v 'udev|sys|run|pts|proc' /dev/md0 on / type ext4 (rw) [root]# [root]# [root]# cat /etc/fstab UUID=4b202d73-d5e4-4678-916c-6220eddb1b60 / ext4 defaults 0 1 [root]# [root]# [root]# grub-probe -t drive / (mduuid/53da4b0e979e6faa68401fe357f506a3) [root]# [root]# [root]# grub-probe -t drive /boot (mduuid/53da4b0e979e6faa68401fe357f506a3) [root]# -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/CAOdo=sysaqr_et+frb7f+tyttsxys7v_tm4jhhfxpc-nivt...@mail.gmail.com
Re: mdadm error - superfluous RAID member
On 13/06/12 04:47 AM, Steve Dowe wrote: Hi, I'm trying to re-use an older server, installing squeeze (6.0.5). I'm using software RAID and LVM on the machine (details below). But I must be doing something wrong with the disk set up stage in the installer, as when it boots I see an error flash up quickly: error: superfluous RAID member (5 found) It appears that the initramfs then gets loaded, the RAID detection fails and it then looks for the LVM volume group, which it can't find (as the LVM group exists on the RAID device). I see this output: Loading, please wait... mdadm: No devices listed in conf file were found. Volume group vgbiff not found Skipping volume group vgbiff Unable to find LVM volume vgbiff/lvroot same messages appear but for lvswap Gave up waiting for root device snip ... It then drops me into the BusyBox shell, with initramfs prompt. I can then activate the RAID simply by doing (initramfs) mdadm --assemble --scan mdadm: /dev/md/0 has been started with 5 drives and 1 spare. and then activate the volume group, using: (initramfs) vgchange -a y 2 logical volume(s) in volume group vgbiff now active Exiting the busybox shell then boots the system. The basic configuration is: - Xeon (64-bit capable) w/4GB RAM - PCI SCSI controller - 6 x 73GB SCSI drives During install, on each drive I created a 500MB primary partition (with /dev/sda1 being for /boot) and then a second partition for Linux s/w RAID (label set to fd). In /dev/md0 I then created a LVM partition, and set up the volume group to contain two volumes - one for swap, and one for /. /dev/md0 is comprised of 5 drives running in RAID5, with one hot spare. During installation, I took pains to wipe all the drives and create all partitions anew. When booted, I checked /etc/default/mdadm. The values INITRDSTART='all' and AUTOSTART=true are both set. I also set VERBOSE=true to give me more output when creating a new initramfs. I checked the contents of /etc/mdadm/mdadm.conf - which seems fine. I then issued update-initramfs -vu, and saw the following: I: mdadm: using configuration file: /etc/mdadm/mdadm.conf I: mdadm: will start all available MD arrays from the initial ramdisk. I: mdadm: use `dpkg-reconfigure --priority=low mdadm` to change this. and the last output before cpio builds the initial ramdisk is Calling hook dmsetup - so, in my limited knowledge, this suggests the drive mapper is incorporated into the initramfs also. When I take a peek into /boot/grub/grub.cfg I see: insmod raid insmod raid5rec insmod mdraid insmod lvm in the 00_header section. I'm running low on ideas now. Re-installing grub doesn't help. Running update-grub simply dumps out many more of those error messages: error: superfluous RAID member (5 found). repeats 17 times So it does point to grub being at fault somewhere, rather than the initrd. Have I missed something blindingly obvious? Thanks again, Steve I prefer to create a separate boot array as RAID1 with extra redundant copies. This circumvents a number of issues between grub, initramfs and mdadm. I don't know if Squeeze can actually boot from a RAID5 array in practice but I find it's not worth the aggravation of trying to make it work. For example, Squeeze has problems with booting from partitioned RAID arrays. After running update-initramfs and update-grub, I find that the UUID for the partitions has been replaced with the UUID for the array, so that the boot fails. This particular problem can be solved by fixing the UUIDs in grub.cfg. I don't use LVM myself so I don't know if the LVM messages are from the same root cause. -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/4fd8dee0.6060...@rogers.com
Re: mdadm error - superfluous RAID member
On Wed, Jun 13, 2012 at 2:41 PM, Gary Dale garyd...@rogers.com wrote: For example, Squeeze has problems with booting from partitioned RAID arrays. After running update-initramfs and update-grub, I find that the UUID for the partitions has been replaced with the UUID for the array, so that the boot fails. This particular problem can be solved by fixing the UUIDs in grub.cfg. grub2 was patched about a year ago to boot from a partitioned mdraid /boot but I don't know whether that change made it into squeeze. -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/CAOdo=sxnaczeqoe5ocptidhm1kiewjdoi8p4cwfyu2knngu...@mail.gmail.com
Re: mdadm error - superfluous RAID member
On 13/06/12 19:07, Tom H wrote: On Wed, Jun 13, 2012 at 12:40 PM, Christofer C. Bell christofer.c.b...@gmail.com wrote: I don't believe you can boot from a striped volume (raid5 being a stripe + parity). I found some instructions that may allow this to work but requires packing a non-standard initrd: http://nil-techno.blogspot.com/2009/02/booting-fakeraid-raid5-linux-half-assed.html grub2 can handle /boot on mdraid raid5 (and possibly dmraid raid5 too). That's ok, my boot partition is /dev/sda1 (500MB) - dedicated to being /boot and nothing else, and all my RAID partitions are /dev/sd*2. I didn't realise grub2 could handle that, though. Thanks. -- Steve Dowe Warp Universal Limited http://warp2.me/sd -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/4fd8f694.4030...@warpuniversal.co.uk
Re: mdadm error - superfluous RAID member
On 13/06/12 19:56, Tom H wrote: On Wed, Jun 13, 2012 at 2:41 PM, Gary Dale garyd...@rogers.com wrote: For example, Squeeze has problems with booting from partitioned RAID arrays. After running update-initramfs and update-grub, I find that the UUID for the partitions has been replaced with the UUID for the array, so that the boot fails. This particular problem can be solved by fixing the UUIDs in grub.cfg. grub2 was patched about a year ago to boot from a partitioned mdraid /boot but I don't know whether that change made it into squeeze. I have just found the GNU grub development mailing list discussion, here: https://lists.gnu.org/archive/html/grub-devel/2012-02/msg3.html Although the symptoms are the same as the Debian bug (http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=610184), I'm not sure whether the causes are. I believe, in my case, the cause is the one discussed in the GNU list, namely that grub couldn't accurately detect whether a partition of the whole disk was assigned for RAID use. In the developer's own words, if you have 64KiB between end of disk and end of partition the metadata is exactly in the same place for either if the whole disks are raided or only partitions. And no field which allows to distinguish those... On that basis, and the fact that grub in squeeze 6.0.5 seemed to exhibit the problem, I decided to update the machine to testing/wheezy instead and see if the problem disappears. I can confirm that it has. The error message no longer appears at boot time and I don't need to intervene to get to my login prompt. For anyone reading this in the same dilemma, I'm not sure if things like this would get backported to squeeze or not - perhaps someone has an idea how to find out... Thanks, Steve -- Steve Dowe Warp Universal Limited http://warp2.me/sd -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/4fd8fa19.9040...@warpuniversal.co.uk
Re: mdadm error - superfluous RAID member
On Wed, Jun 13, 2012 at 4:37 PM, Steve Dowe s...@warpuniversal.co.uk wrote: On 13/06/12 19:56, Tom H wrote: On Wed, Jun 13, 2012 at 2:41 PM, Gary Dale garyd...@rogers.com wrote: For example, Squeeze has problems with booting from partitioned RAID arrays. After running update-initramfs and update-grub, I find that the UUID for the partitions has been replaced with the UUID for the array, so that the boot fails. This particular problem can be solved by fixing the UUIDs in grub.cfg. grub2 was patched about a year ago to boot from a partitioned mdraid /boot but I don't know whether that change made it into squeeze. I have just found the GNU grub development mailing list discussion, here: https://lists.gnu.org/archive/html/grub-devel/2012-02/msg3.html Although the symptoms are the same as the Debian bug (http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=610184), I'm not sure whether the causes are. I believe, in my case, the cause is the one discussed in the GNU list, namely that grub couldn't accurately detect whether a partition of the whole disk was assigned for RAID use. In the developer's own words, if you have 64KiB between end of disk and end of partition the metadata is exactly in the same place for either if the whole disks are raided or only partitions. And no field which allows to distinguish those... On that basis, and the fact that grub in squeeze 6.0.5 seemed to exhibit the problem, I decided to update the machine to testing/wheezy instead and see if the problem disappears. I can confirm that it has. The error message no longer appears at boot time and I don't need to intervene to get to my login prompt. For anyone reading this in the same dilemma, I'm not sure if things like this would get backported to squeeze or not - perhaps someone has an idea how to find out... Since metadata 1.1 or 1.2 stores the metadata at the beginning rather than at the end, perhaps using a partitioned mdraid device with that metada works with squeeze. -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/CAOdo=Sy4ReRPHKdZQw+AbqJwjnhBgfMFXNodQNz-=znopze...@mail.gmail.com
Re: mdadm error - superfluous RAID member
On 13/06/12 23:15, Tom H wrote: Since metadata 1.1 or 1.2 stores the metadata at the beginning rather than at the end, perhaps using a partitioned mdraid device with that metada works with squeeze. Good idea. I'll boot it up with a live CD and report back soon. -- Steve Dowe Warp Universal Limited http://warp2.me/sd -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/4fd915f4.4010...@warpuniversal.co.uk