[ Date Index ] [ Thread Index ] [ <= Previous by date / thread ] [ Next by date / thread => ]
On 15/08/13 20:48, Simon Waters wrote: > On 15 Aug 2013, at 18:42, Mark Evans <mpe@xxxxxxxxxxxxxxxxxxxxxx> wrote: >> I tend to use this or the "If Mark was thinking he backed up the >> /etc/ssh first" method. > Good idea, but things are more dynamic in the new work environment because we do a > lot of testing, experimental builds, and things are getting more dynamic. My > colleague will create a new server just to check out an alleged bug or oddity in > an OS, rather than mess with existing systems. > > We need to get better at this as well. It was taking 2 hours to build a 3 server > test environment with all software and services installed and configured. We are > now using snapshots to reduce that. But we need to get smarter since we have no > template capability in the snapshots, and it would be desirable to build them from > clean OS each time as it is easier to apply patches and keep clean that way, and > proves we still can do it as well - thus proving our build scripts are valid. What virtualization system(s) are you using, and why on earth do you not have multiple VMs already set up *as templates* ready for immediate cloning to a new instance? The only limits I have on deploying new instances is the SAN keeping up as I spawn multiple clones simultaneously. I also don't see why it's desirable to build from a clean OS each time either, apart from proving you are capable and your build scripts work, as you said (fair enough). But it's a massive waste of time. Build, script, template and drop the master image into your repo once and once only. Work incrementally on your masters as required: WSUS/Puppet/Chef/etc to redo from scratch what you know already works is... let's just say wasteful. Please don't construe this as being critical* - after all, I know nothing of your working environment, in-house rules, guidelines and so on: there are no doubt perfectly sensible reasons for any restrictions you have to deal with, often with a PHB attached. But when it's taking 2 hours to build a 3 server test rig, you're doing something very, very wrong. My lowly i7 workstation can spin up 3 new VMs and deploy them in mere minutes, the 64 core Supermicro server in the garage can do it in under a minute and the big bucks hardware my employers typically own can do it almost as fast as I type. Admin overhead post-cloning is taken care of via Puppet normally, another blisteringly quick operation. Worst case scenario means ssh-ing in and pulling the relevant scripts from version control (either sh or powershell, depending on OS flavour), or sometimes in the case of a completely new system, actually writing them. I'm only so efficient because I'm so lazy. I automate everything in sight and reuse stuff constantly because every second I redo some task I've previously completed is time I could be drinking beer instead :] Regards * No really: I know you're not an idiot and am curious as to what kind of admin barriers you're bumping up against -- The Mailing List for the Devon & Cornwall LUG http://mailman.dclug.org.uk/listinfo/list FAQ: http://www.dcglug.org.uk/listfaq