1) I would use soft-mirror: During install dedicate s7 to metadb (~10MB is plenty)
cat /etc/lvm/md.tab /dev/md/dsk/d0 -m /dev/md/dsk/d10 /dev/md/dsk/d10 1 1 /dev/dsk/c0d0s0 /dev/md/dsk/d20 1 1 /dev/dsk/c0d1s0 # metadb -a -c 3 /dev/dsk/c0d0s7 /dev/dsk/c0d1s7 # /sbin/installgrub /boot/grub/stage1 /boot/grub/stage2 /dev/dsk/c0d1s0 # metainit d10 # metainit d0 # metaroot d0 # reboot # metattach d0 d20 Watch i syncing: # metastat *Note!* Without monitoring, redundancy is nothing but delayed desaster. It is useless unless you know a redundant component has failed, and you replace and fix it in time - no matter whether it is soft or hard raid, raid0, raid5, raidz or raidz2! Periodically running metastat and look for fgrep 'State:' | fgrep -v 'State: Okay' and trigger a mail or alert is a good idea. Perl is handy for such tasks.. When it happens: a) replace the disk and b) metareplace -e d0 dx0 where dx0 is the disk metastat complaied about, and watch i resync with: c) metastat 2) Well, controllers does not break anything near as often as disks; but it may happen. So that is up to your cost/benefit judgment. 3) Forget PCI-Express -- if you have a free PCI-X (or PCI)-slot. Supermicro AOC-SAT2-MV8 (PCI-X cards are (usually) plain-PCI-compatible; and this one is). It has 8 ports, is natively plug-and-play-suported and does not cost more than twice a si3132, and costs only a fraction of other >2-port-cards where you pay for raid-chip-sets you don't need or even can't use.. si3132 may be an option but I can't recommend it. SAS-controllers are in another league I think; 3-5 times the price of AOC-SAT2-MV8. This message posted from opensolaris.org _______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
