[1570] in SIPB-AFS-requests
Re: Disk
daemon@ATHENA.MIT.EDU (Greg Hudson)
Fri Nov 4 15:35:36 1994
To: Matt Braun <mhbraun@MIT.EDU>
Cc: star-maintainers@MIT.EDU
In-Reply-To: Your message of "Fri, 04 Nov 1994 14:20:59 EST."
<199411041921.OAA21810@medic.MIT.EDU>
Date: Fri, 04 Nov 1994 15:34:46 -0500
From: Greg Hudson <ghudson@MIT.EDU>
> I have a 2 G disk for SIPB>...where will it go ?
I propose we wait until we have the 4GB disk as well, and do the AFS
cutover. At the point of the AFS cutover (which can happen before the
RTFM cutover), we will have:
Temporary use of ops-temp-w20
Permanent use of new-rosebud (== old-beacon)
A 4GB disk, a 2GB disk, and a 1GB disk for permanent use
(The 1GB disk is what's currently sitting on old-beacon.) We can put
the 4GB disk on new-rosebud, and the 2GB and 1GB disk on ops-temp-w20.
We then move the 2.5GB of data on rosebud to ops-temp-w20, bring
new-rosebud online, and move the data onto new-rosebud.
Once this is done, we will have an extra 1GB of AFS space (plus the
expected .5GB). Assume that we also do the RTFM migration, leaving us
with a 2GB disk and three 1GB disks (one from rosebud, one from
old-RTFM, one left over from the transfer).
We put the three 1GB disks on new-bed and do a news cutover, leaving
us with old-bed's two 1GB disks. We put one of those on beacon, and
are left with a 2GB and a 1GB disk. We put both of those on
new-rosebud, giving us the other 3GB of disk. (We can put the 2GB on
new-rosebud independently of the news cutover.)
Does this make sense to anyone? Just in case it did, I will confuse
any such people with:
Plan B
------
We put the 2GB disk on bed and cut it over, leaving us with bed's two
1GB disks. We use those 1GB disks as swap during the AFS cutover, and
then at some point add another gig to bed.
This allows us to do the news cutover sooner, but assumes that we want
to leave bed with a 2GB and 1GB disk, instead of three 1GB disks.
Currently, I think the news maintainers favor three separate spindles,
but keep in mind that a 2GB disk will be faster, and we only have one
SCSI bus. It's not clear which option is better.
