On Tue, 2005-03-29 at 11:17 -0500, Jim Bartus wrote: > I'm interested in how you guys would approach this problem. > > Say I have two webserver clusters. One www.foo.com and one for > www.bar.com. Each with say 3 webservers and an app server. How would > you go about: > > 1.) Upgrading a package across all webservers? (say an apache security fix) > > 2.) Upgrading a package on only the www.bar.com webservers? (say I want > to test out a new apache version on one first) > > 3.) Upgrade a package (say the kernel) on all www.bar.com servers, but > no www.foo.com servers. > > Basically what I want to be able to do is have an overlapping group > structure so that I can granularly decide to update a package on > "machines in the foo group that are also in the webservers group". you might want to make a yumgroups file with groups like: foo-webserver: foo-appserver: foo-testing: foo-servers: bar-webserver: bar-appserver: bar-testing: bar-servers: webservers: appservers: all: then make sure you're scripts are only ever running a: yum groupupdate whichever of the above groups. then only list the packages in each of those groups that you want to have on those systems. option number 2 - run a couple of local repos and only populate the ones you want. -sv
