Re: Install and RAID
On Tue, 23 Jan 2001, Michael Stone wrote:
> On Mon, Jan 22, 2001 at 04:18:21PM -0600, Steve Langasek wrote:
> > RAID is far less important, in my mind, than a journaling fs. The purpose of
> > RAID is to ensure that hardware failure doesn't result in downtime. But in my
> > experience, your average "workstation for a person who does important work"
> > suffers much less downtime and data loss from hard drive detonation than from
> > unexpected power outages. Many of the problems RAID is intended to solve
> > can also be dealt with by throwing more hardware at the problem (SCSI disks
> > with lower failure rates and higher access speeds, or hardware RAID
> > controllers). OTOH, there's no hardware solution that solves the problems
> > posed by the lack of a journaling filesystem.
> I think you're off base on the purpose of software raid and of
> journalling filesystems. Even if you don't have an uptime requirement,
> you might still want your data--and the cost of a second hd is
> significantly less than the cost of any backup solution I'm aware of.
> No, software raid doesn't help with accidental deletion or fs
> corruption, but I've lost enough hd's to want insurance against that
> problem. As for jfs's, they do *not* prevent data loss from power
> failures, they merely avoid the necessity of fscking on reboot. Any data
> not finally committed to the disk *will be lost* on power failure, jfs
> or not. If that's a problem for you, get a ups.
But if all you're worried about is having a backup of your data, then you
could use any number of other methods to mirror data from one drive to the
other and use that as a backup solution; there's no need for using software
RAID1. The advantage of RAID1 is precisely that you don't suffer downtime in
the event of a hardware failure.
And you're right that a jfs doesn't completely prevent data loss from power
failures. It does, however, prevent precisely the sort of data loss that fsck
looks for.
> That said, I'd support raid as an option, but not as a default.
On this, we agree. :)
Steve Langasek
postmodern programmer
Reply to: