[3228] in SIPB-AFS-requests
Re: migrating rosebud to a Sun
daemon@ATHENA.MIT.EDU (Mike Whitson)
Sat Jan 16 02:58:36 1999
To: sipb-afsreq@MIT.EDU
Cc: zacheiss@MIT.EDU
From: Mike Whitson <mwhitson@MIT.EDU>
Date: 16 Jan 1999 02:58:27 -0500
In-Reply-To: mhpower@MIT.EDU's message of Mon, 11 Jan 1999 14:46:27 -0500
Matt: I just read and reread your mail three times, and I find I'm
still confused as to what you're proposing we do. Let me see if I
understand:
You feel we should:
- keep all of the servers' /os and /srvd local on the servers,
based on a canonical tree in the sipb cell
- Buy a new tape drive and make regular UFS backups of the
servers' system disks
- Use these UFS backups instead of (or in addition to)
reinstallation, to restore the servers in case of a
crash
The benefit you cite compared to the ASO model (mostly remote system
packs and mkserv) is:
- It requires less specialized knowledge than the ASO model.
The benefit you cite compared to the ASO dialup model (local system
packs which are recreated from AFS at install time) is:
- It takes less time.
Do I understand correctly, or am I missing something?
Assuming for the moment that I understood you (please disregard this
if I'm merely confused about what you're suggesting), I can't agree
with your proposal.
* I don't believe anybody is suggesting that we should use the
"dialup" model, so I believe the factor of install time is moot.
Recreating a server using the ASO instructions and scripts, in my
opinion, takes equal or less time than recovering the system disk from
tape backup would.
* I will contend that the set of instructions and scripts ASO
uses for creating an AFS server are clear enough that they require as
much or less specialized knowledge than restoring the system disk from
tape backup. If they aren't, having the feedback such that we can
make them more clear would be valuable to ASO -- having the procedure
be as straightforward as possible is one of ASO's goals.
* To set up this system, we would need to go to a lot of
trouble, including setting up backup software and scripts, possibly
buying a tape drive, and developing instructions and procedures such
that someone with only general unix administration clue could perform
them. Given that I don't believe this method has significant
advantage over ASO's model, I don't think this would be effort
well-spent.
* While it may be true that historically, SIPB has been slow
to move on server upgrades and the like, I don't believe that because
it has been true in the past means that it is necessarily optimal, nor
does it mean that it necessarily will always be done that way. It
would be nice to have this project done in a timely manner before the
end of IAP, while people still have time away from classes and the
like.
In short, (unsurprisingly) I think ASO's setup for AFS servers will
work just fine here.
-mike
