[2556] in SIPB-AFS-requests
Re: RFD - Backup system: tape retiring, system overhaul/upgrade
daemon@ATHENA.MIT.EDU (Emil Sit)
Sun Nov 24 16:18:14 1996
To: Jonathon Weiss <jweiss@MIT.EDU>
Cc: star-maintainers@MIT.EDU
In-Reply-To: Your message of "Tue, 19 Nov 1996 10:57:20 EST."
<199611191557.KAA26862@tla.MIT.EDU>
Date: Sun, 24 Nov 1996 16:17:16 EST
From: Emil Sit <sit@MIT.EDU>
> [hostlist taken from btc:/usr/local/rmt/.klogin, minus dragons-lair]
> as far as I know bed and beacon do ont use the OSU backup system.
If bed and beacon don't use the osu system, they can be removed
from the list and it can be suitably compressed. (in response to Matt's
point).
> This is for the same reason that we don't just backup all of the
> servers at once onto a DLT cartridge. It requires dependencies in
> root access to various servers that SIPB has not been able to agree
> upon.
having spent about 30 minutes poking around the scripts, I think it is
possible for us to have a single (RCS maintained!) location with
all the scripts and yet not have to worry about root access problems.
all-backups rsh's over to the host on which backups are to be done on
to run do-backups. It is a simple patch to have it rsh -l root
using the current users root tickets. That should make the issue
of root access go away, right?
Under this system, the database would be stored in AFS, with the
maintainers and rcmd.{a-c,charon,etc..} on the ACLs. (there may be some
byte-order issues, but the docs say they can be fixed) There appears to
be some AFS token-getting code in do-backups on anxiety-closet already.
A single machine would be picked to run-clean-db nightly and such. The
tapes would all use one set of numbers and the system would rotate
between all of them for all hosts (I'm not sure about that point, but
that sounds right.)
In terms of DLT, if there's any one with root on all four machines
(mhpower?), that person could certainly start a dump for all four hosts.
It shouldn't be too hard to configure the system to support the DLT
drive as well.
How's that sound? Integrating the backup system would require some
amount of time, which I might not have much of in the immediate
future, however. I'd certainly be able to do it over IAP. (And
at that time, I had this strange idea to rewrite the scripts in
Perl5....)
Emil
--
Emil Sit / Bronx Science '95, MIT '99 -- ESG, SIPB.
Email: sit@mit.edu / Web: http://web.mit.edu/sit/www/
PGP KeyID: 0xE63561E9 / Fingerprint: A68FD0693EDABA19 2671EC1F22498F58
