[8416] in Info-AFS_Redistribution
RE: Replication volume swaps
daemon@ATHENA.MIT.EDU (Jeffrey Hutzelman)
Sun Jan 21 18:59:06 2001
Date: Sun, 21 Jan 2001 18:10:14 -0500 (EST)
From: Jeffrey Hutzelman <jhutz@cmu.edu>
To: "Neulinger, Nathan R." <nneul@umr.edu>
Cc: "'Steve Lammert'" <shl@mars-systems.com>,
Jeff Blaine <jblaine@linus.mitre.org>,
Mitch Collinsworth <mitch@ccmr.cornell.edu>, info-afs@transarc.com
In-Reply-To: <F349BC4F5799D411ACE100D0B706B3BB1D3901@umr-mail03.cc.umr.edu>
Message-Id: <Pine.LNX.4.21L.0101211808160.3147-100000@gtok.andrew.cmu.edu>
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
On Fri, 29 Dec 2000, Neulinger, Nathan R. wrote:
> Uh. The below is a really bad idea!
>
> During the time between the first and completion of the third step, your
> users won't be able to access the data reasonably, since the readonly volume
> won't exist any more in the vldb. Always create the new one first, then
> remove the old one.
Correct. We used to do addsite, release, remsite, (wait 2 hours), zap.
This is because some very old clients would take up to 2 hours to notice
the VLDB change and deal correctly. Modern clients recheck the VLDB any
time a fileserver tells them it doesn't have the volume, so it's not
longer necessary to do that -- and since we retired our Mach systems,
we've done addsite, release, remove.
-- Jeffrey T. Hutzelman (N3NHS) <jhutz+@cmu.edu>
Sr. Research Systems Programmer
School of Computer Science - Research Computing Facility
Carnegie Mellon University - Pittsburgh, PA
