[1539] in SIPB-AFS-requests

home help back first fref pref prev next nref lref last post

backups

daemon@ATHENA.MIT.EDU (yoav@MIT.EDU)
Wed Oct 19 12:49:35 1994

From: yoav@MIT.EDU
Date: Wed, 19 Oct 94 12:48:55 -0400
To: star-maintainers@MIT.EDU


okay.. for those whoi missed the maintainers meeting, it was proposed that until we
have a second tape drive, we should come up with schedules to avoid collisions.

I will throw a random schedule here, and anybody who disagrees may feel free to
point this out:

	each day lasts from about noon to about three A.M.  between three A.M. to
noon, no tape in the drive should be punted (hopefully this is never done anyway),
and that time is sort of reserved for finishing off REALLY long backups..


	Monday  	-	AFS backups
	Tuesday		-	server backups
	Wednesday	-	user backups and AFS backups (AFS should be polite but
							has prioirity)
	Thursday	-	user backups and server backups
	Friday		-	first come first served backups (AFS,server, and user).
	Saturday	-	AFS backups
	Sunday		-	server backups

This would mean we would reccomend users came in on Friday if they can, or Wed. or Thu.
if they can't.. really discourage them from Mon, Tue, Sat. and Sun. (though I'm sure
we can deal with a user taking the drive for about an hour).

	Any suggestions/objections?
	(we should announce this to sipb-all if we agree on a set schedule.)



	SECONDLY:
		I'm not really into this schedule thing, and was wondering if
	people really couldn't deal with the current method?  If there are 
	explicit collisions, I believe AFS should have priority (server backups
	are RARELY used, and most of the stuff is replicated in AFS anyway).
		If there are too many cases of busy tapes just when you planned
	on doing backups, then we can revert to this. As far as I noticed, most
	of the time, the tape drive is unused.

	THIRD:
		If this really is a problem, we should be more serious about getting
	another drive. (and I still think getting a DAT wuold make sense.. use it
	for server backups, and if both DAT drive and machine blows, work a little
	more at restoring the server from AFS (which is waht should be done in the
	first place anyway, and the likelihood of that happening is REALLY low)..


			yoav

home help back first fref pref prev next nref lref last post