On 27/1/2013 5:11 μμ, Johnny Hughes wrote: > So upgrading one package can cause a domino effect that means you have > either broken a bunch of packages or you have to rebuild a bunch of > packages. That is why you should *only* upgrade what is VITAL for your application(s), and in a carefully planned manner, as Johnny explains. Building on the example I gave earlier, OpenLDAP Server, which IS important to be current in production due to critical bug (not only security) fixes, can be safely upgraded using LTB project RPMs (http://ltb-project.org/wiki/) which allows a current version WITHOUT affecting the rest of the system (by using alternate paths to install libs, executables etc.). Any other Servers in need for a current version can be safely upgraded through a careful approach to make sure they don't break other things. So, get to know your OS, your apps, your (S)RPMs; then plan, test on a non-production system, then plan deployment in production. For example, we *only* upgrade other critical apps like Postfix, Dovecot etc. when running as enterprise servers. There is no other option here; for example, Postfix as available in RHEL/CentOS 5, is no more supported by the Postfix developer. We have to build our own RPMs based on our particular requirements. We enjoy the stability of CentOS because it allows us to only upgrade what is vital. We don't care about Gnome and other modernization(s). Our CentOS 5 enterprise servers do not even have Gnome installed. Nick