[2662] in SIPB-AFS-requests
New server should be ready
daemon@ATHENA.MIT.EDU (Greg Hudson)
Sat Feb 15 13:07:55 1997
Date: Sat, 15 Feb 1997 13:07:28 -0500
From: Greg Hudson <ghudson@MIT.EDU>
To: sipb-afsreq@MIT.EDU
Okay, I reinstalled the new server yesterday and set up the AFS
software today. The disks appear to work fine. A few quirks I ran
into:
* The ops install points /usr/include at /srvd/usr/include,
and we break the /srvd symlink; you have to copy in
/usr/include from the ops packs before the kernel will build
properly.
* If you tar over all of /usr/afs from another machine, afs2sl
will syslog all of the log messages. Better to tar up the
directories separately, I guess, and omit /usr/afs/logs and
/usr/afs/local.
* I commented out the synclocker and release scripts from
crontab today; they will have to be uncommented when we put
the server into production.
I would like to let other people look over the machine to see if
anything looks wrong. If not, we can cut over from ronald-ann to the
new machine, say, next Saturday night. For reference, here is the
plan I proposed in [2587] to do the cutover:
* Install the machine as sipb-server-1 (lots of installation
details glossed over, of course) and bring it up as an
AFS file server.
* vos move all the volumes from ronald-ann to sipb-server-1.
* Bring ronald-ann and sipb-server-1 down; bring up the new
machine as ronald-ann.
* vos syncvldb to point the VLDB at the volumes on ronald-ann.
* Start up a db server on ronald-ann. It should take a dump
from one of the other servers and everything should work.
If nobody finds anything wrong, I move the volumes from ronald-ann to
sipb-server-1 sometime this week.
