hi andree !, Le Samedi 29 Avril 2006 09:48, Andree Leidenfrost a écrit : > First, I noticed your patch explicitly zeroes the superblock in a > separate step before creating the RAID arrays. I have found that 'mdadm > --create --force --run' will ignore existing superblocks. Am I missing > anything here? > > Second, you provide scripting for updating the uuid in intird images. I > can restore and on reboot the RAID arrays are started automatically > (using a stock 2.6.16 kernel), which is what I would expect. What is the > uuid update needed for? >
I don't remember exactly, but I think both should address the same issue : the --zero-superblock step was a lame attempt to fix the issue that is actually fixed by the initrd stuff. my config was : stock debian sarge, stock debian kernel 2.6. boot partition is lvm-on-(soft)-raid1. in this configuation, the RAID array refused to boot because the uuid is explicit and hard-encoded in the mdrun step. 'script' file in initrd.img looks like : ROOT=/dev/md0 mdadm -A /dev/md0 -R -u 6eb547bc:9be35f35:5037390f:1934535e /dev/hda1 /dev/hdc1 after a mondo restore, these uuid will have changed... and the array won't run :( if you doesn't need this step, maybe it's because : - your boot partition is not in an array ; or - you tried your restore on the same array (ie, same uuid) ; or - i've totaly misundestood what's happening in there ;) ! hope this will help ! i'm looking forward to your patches ! :) cheers, guillaume
