[533] in SIPB-AFS-requests
SIPB Locker in the SIPB cell..
daemon@ATHENA.MIT.EDU (Derek Atkins)
Tue Nov 12 18:33:48 1991
To: sipb-afsreq@Athena.MIT.EDU, probe@Athena.MIT.EDU
Cc: warlord@Athena.MIT.EDU
Date: Tue, 12 Nov 91 18:33:14 EST
From: Derek Atkins <warlord@Athena.MIT.EDU>
When I was testing a new copy of the sync-code today on ronald-ann,
it majorly hosed the whole cell! The major reason is that while
the script was being run on ronald-ann, the read/write copy of the
locker is on rosebud, so the load gets shifted to BOTH machines!!!
While this script was running the synctree, the load on ronald-ann
went to about 4... (I didn't check the load on rosebud)
I think that a solution to this problem would be to move the read-write
copy of the volume onto ronald-ann, and having a replication site
on rosebud as well. This would solve a few things:
1) it will give up a backup of the volume
2) it will allow the sync script to TOTALLY use ronald-ann,
leaving rosebud to serve the users, and ronald-ann to be
used to sync from.
Are there other solutions? What exactly ARE the problems? I'd
like to get people's opinions on this... (Please no 'lets move it
back into the Athena cell')
-derek
