On Tue, 2004-10-12 at 09:25 -0500, John Foust wrote:
Has anyone mentioned that if the OS or hardware or
horrid software goes
astray and writes crap all over the disk, RAID doesn't save you
but the 'periodic rsync' approach does?
It's only as good as any other backup strategy though; it'll only give
you live data up until the last rsync was done. RAID's more for keeping
critical data on-line even in the event of failure.
As I said earlier, only Jay can decide what hardware and what level of
redundancy is needed.
However, *if* the websites hosted on the classicmp server are being done
so for free, I'd be tempted to say that they're not under any kind of
support deal or service guarantee, *and* that the authors would be
expected to have their own backups of their particular site. In other
words, put that data in a non-RAID environment, and if a disk blows up
then it'll knock the site(s) out until another disk is brought on-line
and the data either restored from Jay's own backup or from indivial
backups of the site owners.
The mailing list (including the list archives) is another matter - there
I'd be inclined to say that system failsafe ability and uptime is more
important, and that the data there is more critical. This is
particularly true for the mailing list config and archives - if the list
goes down for a short while it's less of an issue. This is all assuming
that the mailing list is the primary function of the classiccmp server,
of course...
Splitting the problem into two this way might be a sensible choice,
though. Small amount of RAID storage for mailing list functions and
application data, and a large amount of non-RAID disk for other data
(such as bitsavers mirror).
cheers
Jules