[1606] in SIPB-AFS-requests
Re: Cutover plans
daemon@ATHENA.MIT.EDU (Greg Hudson)
Tue Nov 15 20:52:48 1994
To: mhbraun@MIT.EDU
Cc: Derek Atkins <warlord@MIT.EDU>, Greg Hudson <ghudson@MIT.EDU>,
sipb-afsreq@MIT.EDU, yandros@MIT.EDU
In-Reply-To: Your message of "Tue, 15 Nov 1994 19:35:45 EST."
<199411160035.TAA25549@medic.MIT.EDU>
Date: Tue, 15 Nov 1994 20:49:46 EST
From: Greg Hudson <ghudson@MIT.EDU>
> 1) outages....Greg's way has no user visible outage and if we have
> the swap space it probably is the best.
Derek's plan doesn't involve any outages. It was (I'm just quoting):
1) put the new machine up as new-rosebud, setup as a
fileserver only.
2) start moving all the RW data off rosebud and onto new-rosebud
directly.
3) re-add the RO volumes to new-rosebud
4) Once all the volumes have been moved off rosebud onto
new-rosebud, make sure that no volumes remain on rosebud.
5) add new-rosebud as a db server to the other machines and
then create the db processes on new rosebud
6) remove old-rosebud, if we want, from the db list.
Step 6 might cause some problems, but we don't need to do that. As
long as the name of the new server is something like "new-rosebud", we
will not need to coordinate with IS at all.
It also only requires one Delni slot, which we have free already.
When we do the news cutover, we will have a second Delni slot free for
ronald-ann to move.
> 4) Doing everything at once changes many varaibles. It is not clear
> we want to change that many.
Derek's plan doesn't do everything at once; it just handles the
migration without swap.
I don't really know enough about the details of AFS to know how we're
going to coordinate AFS changes with IP address changes of the names
ronald-ann and rosebud, but I don't think this will make it
significantly harder.
