From: Les S. <le...@cy...> - 2007-12-05 12:44:49
|
Matthew Metzger wrote: > thanks for the response. Since this will be for BackupPC, all of the > data pool has to be on one volume. If I ever did run out of space, > having LVM could be a temporary life saver. From what I've read, it > can't hurt any to use LVM. > > Your case for RAID 5 is very convincing. I think that I may change to > that with LVM layered on top. If something goes wrong with one of my > disks, how will I know? And more importantly, how will I know which one > to replace (the disks are all physically identical). > > in software raid - /etc/mdadm.conf (on redhat based systems), set MAILADDR=you...@yo... it will email you when there is a failure. mdadm --detail /dev/md0 will tell you which partitions make up the array, which are active, failed etc. generally, motherboard sata ports 1,2,3,4 correspond to sda,sdb,sdc,sdd also you should enable, if supported, smartd. it will monitor your disks and email you if there is a problem (/etc/smartd.conf). as for raid 5 ..... interesting comments from John, i've never hit "write hole" problems, but that information is worth knowing, thanks John. If you are expecting a huge load on the server you should probably look at a proper hardware controller with write cache and battery backup. In a perfect world i would always use that, and i'd always prefer raid 1+0 anyway, but then that costs more and not every client justify the additional cost. ;) however you said you had 4 disks and its just for backuppc. so i imagine that if you use software raid5 and did have a disk failure and you put in a new disk to rebuild you could probably disable backups for the duration of the rebuild which may take a few hours. Mind you Linux multi-tasks pretty well ;) Regards, Les |