Actually there's a private discussion already running relating to this,
although I got into it from a different perspective (hardware upgrades).
On Wed, 22 Jun 2005, Ian Stokes-Rees wrote:
> 1. Have two sets of head nodes.
...
> 3. When an update requiring a reboot or some dramatic change which makes
> old WNs incompatible with the new system, take down the "secondary" head
> nodes (once they finish their active jobs), and install the new software
> on them.
> 4. Now as WNs finish their active jobs they can be taken offline,
> re-installed, and re-booted but now under control of the "secondary"
> head nodes.
No quite sure what you mean here - is that not going to confuse the hell
out of everything, as you'll be trying to run two CEs with the same
certificate?
My view is that you really want simply to let each site have many
conventional CEs, then once the "new" CE is going you would transfer the
workers as you suggested.
The difference between our schemes being that I don't buy the two-way
split implied in yours - big sites with disparate groups of nodes might be
cycling round between 3 or more CEs during upgrades.
On the other hand the trouble with the many CE route is that while it's
apparently technically possible, no-one is yet quite sure quite how to
deal with accounting and SFT records, and such.
Hth
Henry
--
Dr. Henry Nebrensky [log in to unmask]
http://people.brunel.ac.uk/~eesrjjn
"The opossum is a very sophisticated animal.
It doesn't even get up until 5 or 6 p.m."
|