> Which is why I think there should still be some manually-managed servers remaining (possibly something like testing). All operations during development are c

> Which is why I think there should still be some manually-managed servers remaining (possibly something like testing). All operations during development are carried out manually, unless already supported by automation (and until, of course). When automating the new operations, a sysadmin must also conduct the steps manually, and also take care of considerations a developer might not have been aware of. > What I didn't see mentioned is that when the automation fails for whatever reason, typically it is a five-alarm situation where minutes count. Generally, this is easily-handled by redundancy, HA, and snapshot rollbacks. Of course anything that's a potential five-alarm situation has redundancy, right? :P