TFS & GO & Chef, oh my: Part 12 - Conclusion, and lessons learned


Well, the presentation went over really well, so this project is now a real thing.  I'll probably post a follow-up once the pilot is in place that will highlight the questions that needed answering before it could go ahead.

If you're looking at having to go this route, some questions to consider:

  1. Why are YOU looking at solving this problem?  If it's just you, either you or everyone is doing it wrong.
  2. Is there a cross-dept team linking the project together?  Why not?
  3. What is 'DevOps' to you?  It should mean 'making things flow better'.
  4. (realizing the irony of this) Have you already picked out tools?  If so, you're doing it wrong.
  5. Do you understand exactly what you're trying to accomplish?  I mean, really?  You sure?

Feel free to reach out if you have any questions.  I know these posts have been incoherent at times.

Lessons learned

  • Technical: Using Go & Go agents to remote around means you're always kinda wondering what user you are, and why such and such doesn't work - truly highlights how little you grasp of authentication (or how well Microsoft has abstracted that process from users)
  • Technical: Chef documentation is still a work in progress
  • Technical: Missing idempotency features in key cookbooks (IIS, I'm looking at you!)
  • Technical: No real simple explanation of how to layer/architect your cookbooks/recipes/roles - took hours and hours of reading to figure out (and still not certain)
  • Technical: Having to learn ssh keys, how Linux deals with sessions coming from Windows, etc.
  • Technical: (Still) learning to use Git
  • Technical: Trying to figure out the best way to deal with multi-environment configs in IIS (and the config source usage gotcha - must be inside the dir!!! i.e. cannot be a separately called dir, aka repo)
  • Technical/Process: Trying to reverse engineer MSI packages built by our company so long ago nobody knew anything about them
  • Technical: Learning to be a developer with no training (no choice here - in the coming years, Ops must learn to code)
  • Process: Trying to understand our deployment process with the mentality that it was an existing 'good' thing (when it was a very very bad thing, i.e. trying to automate a crappy process when you have the option of re-doing the process instead)
  • People: Trying to do this without (dedicated) input from all teams is bad.  In our case we had no choice, but that didn't make it good.
  • Process: Trying to do this without FULLY grasping the scope of the project was a bad idea.  Again, we had no real choice here, but yeah.  Bad.
  • People: The reason why something has never been done before is almost always NOT technical...sometimes it just takes someone who is willing to do it.
  • People: When faced with the choice of 'stay the same and decay' or 'invest time to move forward' - people rarely will rationally choose the former.  Conversely, they will rarely choose the latter voluntarily.  (just my experience)

Comments

Popular posts from this blog

DFSR - eventid 4312 - replication just won't work

Fixing duplicate SPNs (service principal name)

Logstash to Nagios - alerting based on Windows Event ID