Staging

Stage updates before deploying them. In cpacman world that means - you fetch updates from upstream:

 # up2dater.py -h site.x.com

most probably you're going to be doing it in batch:

 # for s in s1 s2 s3 s4 ; do up2dater.py -h ${s}.mysite.com ; done

Better yet use find_server.py to do the dirty job of finding appropriate servers:

 # for s `find_server.py -N -d` ; do up2dater.py -h ${s}.mysite.com ; done

After that you only use update.py to install errata's on servers of appropriate waves:

 # for s in `find_server.py -N -d -m exact -s priorities update_priority=1`; \
   do update.py -h ${s} ; done

And once done with that - push those to servers as usual:

 # for s in `find_server.py -N -d -m exact -s priorities update_priority=1`;\
   do installer.py -h ${s} -u root ; done

However batching thinks like in last step might be dangerous because if you do have some problem RPM is not smart enough sometimes to return error code so cpacman has no idea that install failed and confirm will do stupidest thing ever. Thus confirm.py is called manually for each machine once you DO confirm that everything is OK.