On Sunday 07 March 2004 17:56, Erich Waelde wrote: [snip] > I managed to set up a /target tree on lvm on raid1 on scsi, but base > install failed (could not download coreutils), and no smoke test has taken > place. See below for details. > > To summarize: I found all the neccessary tools! Good work! Thank you. > It needs a bit of polishing here and there, but AFAICS nothing really > difficult.
I'll try to comment as good as I can on the problems. [snip] > * Load installer components > . lvmcfg, nobootloader > > 1. I did expect sw-raid or md to show up here, but it didn't As I have chosen to load it as default, it is not loadable. I haven't quite understood how the installer components list is generated, but I will play a bit with it soon. > * Detect network hardware > * Configure the network (dhcp) > * Detect hardware (disks: 2 scsi, 1 ide, cdrom ide) > * Partition a hard drive > ide0 part1 30 MB /boot > scsi0 part1 1 GB + > scsi1 part1 1 GB ==> /dev/md/0 > scsi0 part2 1 GB + > scsi1 part2 1 GB ==> /dev/md/1 > > * configure md > > 2. "configure md" is listed _after_ "configure lvm", where I would expect > it the other way round. Not a big deal. Hehe, It's an easy fix too, I'll add it in my next image. > 3. of course I forgot to mark partition type 'fd' and hat to go back. > Precise error message made that easy, though. :-) > > 4. it took several attempts until I found that mdadm (cat /proc/mdstat) > had some old junk information (2 broken md devs), which would prevent > the new partitions from showing up in the menus. > > # mdadm --stop /dev/md/0 > # mdadm --stop /dev/md/1 > fixed that. > > Of course I zeroed out the drives first and rebooted several times > after changes in the partition tables. But heck. I had this problem a couple of times too. However I don't think there is an easy way to fix this, since broken superblocks shouldn't occour in general (or at least we can't assume that existing superblocks are broken). > 5. the first raid device was created and started to sync. Then I wanted > to create the second device. At first it seemed that it would block on > the sync to finish (I used to raidtools2, rather). But that proved > wrong. I could kill the "mdadm --create ..." process and restart. > Again it did block (blue screen, nothing moving). So I killed it again > and issued the command on the shell: > > # mdadm --create /dev/md/1 --force -l 1 -n 2 -x 0 > /dev/scsi/...target{0,1}.../part2 > > which did ask me, whether I wanted to proceed, yes or no. So maybe it > needs a "do-it-and-don't-bother-me" switch set? > > At least I got 2 raid1 devices, synced and up, both consisting of 2 > partitions on the scsi disks. Hmm... The mdadm --create thingie blocked, right? What was it complaining about, existing superblock or size warnings? I have had this issue too, and --force isn't as forcefull as I would want it to be. Maybe I should patch mdadm with an --force-for-real argument ;-) > * configure lvm > > 6. lvmcfg complained, that it would not find any usable partitions. Which > is true, kind of, since none where marked 'LVM'. However, it's wrong, > kinf of, too, since it doesn't consult /dev/md devices, it seems. > > # pvcreate -ff /dev/md/0 > # pvcreate -ff /dev/md/1 > > Unfortunately, no change (I reselected "configure lvm" from the menu). > So I went ahead to do it on the shell > > # vgcreate vg00 /dev/md/0 > # vgextend vg00 /dev/md/1 > > which worked fine. Sounds like lvmcfg needs some work... I'll look into it at a later stage, when the rest works a bit better. [snip] > * Install the base system > > fails consistently "cannot download coreutils". Did delete the file in > /target/var/cache/apt/archives, but no change. But I believe this is a > problem with my CDRom not liking CD-RW's rather than anything else. Now this sounds strange, haven't touched coreutils. My bet is on the CD-drive too -- but OTOH I am supposed to say that, am I not? ;-) Thank you for your feedback, -- Paul Fleischer // ProGuy <proguy at proguy dot dk> PGP key fingerprint: 755A 9FB3 F7E4 DB62 8154 C5D6 381B BBCD 7BE1 FF30 -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]