On Fri, 9 Apr 2004 me@heyjay.com wrote: > Hi, > > I've read all the linux raid how-tos, and thought I was ready to take the > plunge on implementing software raid. I've got a fresh install of Debian > Sarge, 2 disks. 3 partitions (/boot, /, swap). So I installed Sarge, > partitioned hda during the installer with regular partitions: > > abba:/usr/src# sfdisk -l /dev/hda > > Disk /dev/hda: 2491 cylinders, 255 heads, 63 sectors/track > Units = cylinders of 8225280 bytes, blocks of 1024 bytes, counting from 0 > > Device Boot Start End #cyls #blocks Id System > /dev/hda1 * 0+ 11 12- 96358+ 83 Linux > /dev/hda2 12 133 122 979965 82 Linux swap > /dev/hda3 134 1957 1824 14651280 83 Linux > /dev/hda4 0 - 0 0 0 Empty > > I wanted to do a raid 1. So, then I went and partitioned hdc like: > > abba:/usr/src# sfdisk -l /dev/hdc > > Disk /dev/hdc: 1870 cylinders, 255 heads, 63 sectors/track > Units = cylinders of 8225280 bytes, blocks of 1024 bytes, counting from 0 > > Device Boot Start End #cyls #blocks Id System > /dev/hdc1 * 0+ 12 13- 104391 fd Linux raid autodetect > /dev/hdc2 13 137 125 1004062+ fd Linux raid autodetect > /dev/hdc3 138 1756 1619 13004617+ fd Linux raid autodetect > /dev/hdc4 0 - 0 0 0 Empty > > So now I'd build my raid set: > > abba:/usr/src# mdadm --create /dev/md0 --level=1 --raid-devices=2 /dev/hda1 > /dev/hdc1 > mdadm: /dev/hda1 appears to contain an ext2fs file system > size=96358K mtime=Thu Apr 8 12:50:04 2004 > mdadm: /dev/hdc1 appears to contain an ext2fs file system > size=96358K mtime=Wed Dec 31 18:00:00 1969 > mdadm: largest drive (/dev/hdc1) exceed size (96256K) by more than 1% > Continue creating array? n > mdadm: create aborted. > > I've google for that "mdadm: largest drive...than 1%" error with no results > > Any help? Firtly I don't use madm, so no direct help there, but I suspect you built the system onto hda then tried to incorporate hdc ... This won't work as it'll potentially overwrite the partitions as it builds the raid1 set. (as far as I'm aware - I think theres some runnery you can do to do with marking the new drive as failed, but I've never tried it) (I suspect your 1% error is madm complaining because you are going to lose more than 1% of your disk space - for raid1, the opposing partitions really need to be the same size, but AIUI the raid partition will be created the size of the smallest partition) It's a royal PITA with Debian, but this is how I do it: Starting with 2 identical disks: Install a bare minimal system onto hda parition as follows: / 256M swap 2 x RAM /usr 2GB /var Rest of disk. (You don't need a /boot - thats only for older BIOSes which can't boot from cylinders > 1024) Once you have installed a bare minimal Debian onto /dev/hda, then partition /dev/hdc as close to hda as possible. Then the fun starts: unswap /dev/hda2 (and edit it out of /etc/fstab), and use whatever to create a raid1 as /dev/md1 using /dev/hda2 and /dev/hdc2. I use /etc/raidtab and mkraid, but you might want to use madm. Manually mkfs this, then mount it under /mnt and copy / into it, as per the HowTo. (cd / ; find . -xdev | cpio -pm /mnt) Then you need to edit /mnt/etc/fstab to make it show that / is under /dev/md1, edit /mnt/etc/lilo.conf to have the right runes in it (raid-extra-boot, etc.) then run lilo -r /mnt. Run cfdisk (or whatever) to change the partition types to 0xfd and reboot. With a bit of luck you'll now have root mounted under /dev/md1 with no swap, and /usr and /var under /dev/hda3 and /dev/hda4 respectively. Use the same procedure now to create md0 from /dev/hda1 and /dev/hdc1 and mkfs it, mount it under /mnt and copy root back to it. Remember to edit /mnt/etc/fstab (/ is now under /dev/md0) and the new lilo.conf and re-run lilo, set the partition types on /dev/hda1, hdc1 to 0xfd and reboot. Now you have /dev/md0 which is root, /dev/md1 which is unused (it will become swap later on) and /dev/hda3 -> /usr and /dev/hda4 -> /var. Next step is to copy /usr into /dev/md1, so re-mkfs /dev/md1, mount it under /mnt, and do the same cd /usr ; find . -xdev | cpio -pm /mnt trick as before, adjust /etc/fstab to mount /usr from /dev/md1 and reboot. Now use madm or whatever to create a raid1 (as /dev/md2) from /dev/hda3 and /dev/hdc3 and mkfs it. (don't forget to change the partition type to 0xfd) Guess what comes next ;-) Copy /usr currently under /dev/md1 into /dev/md2. Edit /etc/fstab and reboot. Do the same trick finally for /var - copy it into md1, adjust /etc/fstab, reboot, create /dev/md4, copy it back again, adjust/etc/fstab, reboot, and finally, mkswap /dev/md1 and swapon /dev/md1 and put that back into /etc/fstab. Then go back into dselect (or tasksel) and finish off installing the packages, etc. you need. You end up wih something like: Filesystem Size Used Avail Use% Mounted on /dev/md0 235M 31M 193M 14% / /dev/md2 1.9G 1.5G 335M 82% /usr /dev/md3 30G 6.9G 21G 24% /var /dev/md4 42G 24G 16G 58% /archive and lion:/# cat /proc/swaps Filename Type Size Used Priority /dev/md1 partition 995896 59092 -1 This is pair of 80GB disks and I have an extra partition here (which started life empty, so didn't need a copy/shunt about thing) Hopefully the next release of Debian will include raid from the install CD... It's otherwise a real PITA to install, but worth it at the end of the day! Good luck... Gordon - To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
