[Linux-cluster] Using clustering to facilitate OS patching

[Mark Wagner]

One of the ways we use clustering is to allow us to patch a box without
extended downtime. I.e., to patch server A (1) relocate services on A to B
(encountering a small downtime), (2) patch/test A with no time pressure,
then (3) relocate the services back to A. Does anybody do the same or
have comments on this patching technique?

We are now beginning to virtualize and using this method in a virtualized
environment seems awkward at best. It is OK for the dom0. The services
are domUs so the relocation step is relocating them. However, if we
run clustering for the domUs we end up running clustering on top of
clustering. Is that reasonable? I can see one pitfall in that if you
aren't careful you could have a clustered domU pair running on the
same dom0.

Mark

-- 
Mark Wagner <markwag at u.washington.edu>
System Administrator, UW Medicine IT Services
206-616-6119