> The real issue here is that we manage machines wrong. The fact that > sysadmins say things like, "if I had more than a few machines I'd set > up Puppet/Chef/CfEngine" should be considered a bug. We should be > using configuration management as the default. Everything should be > done via CM. Software packages should come with plug-ins that expand > the CM verbs/nouns so they can be managed. GUI front-ends should just > manipulate the databases that drive our CM systems. Editing a file in > /etc directly "by hand" should be an obscure art done to teach > internals or to scare children on halloween. Sadly Unix isn't built > like that (today) but that's where we should be aiming.
Many CM advantages have little to do with scale, but everything with: * de-facto documentation (blueprints instead of a diary), tested(!), never kept on the TODO list * predictable changes, because you *know* where you depart from. * preventing "black boxes", servers inherited from colleagues long gone that no one dares to touch * being able to quickly transfer your work in progress to a colleague (without reverse engineering) * never being afraid of being unable to reproduce a service, because of some undocumented change years ago. If a service is important, I'd say the job requires that our colleague or successor is able to reproduce it. Making full backups is sometimes insufficient to recover from a security incident (or to perform OS-upgrades). Sometimes, you have to reinstall from scratch and repeat all configuration steps you ever did to get a service running again. You could have tested your documentation by having all steps repeated by someone else, but who has time for that? And do you recheck, for all those changes after going into production? Am I too demanding, stating these requirements? To be honest, I would *never* apply for any sysadmin job anywhere if CM isn't at least on the wish list. I simply wouldn't feel like I could guarantee a timely recovery or meet expectations without it. And this is from someone who used to install Linux servers with Slackware, rebuilding packages and kernel by hand on every system ;-) Best, Hans NL
_______________________________________________ Tech mailing list Tech@lists.lopsa.org https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech This list provided by the League of Professional System Administrators http://lopsa.org/
