I just noticed it's been 1 month and 5 days, depending on how you slice it, since my initial chef-themed entry. That's several business days, without a complete working result that can build my software. Granted, a lot of learning and bootstrap stuff. But, I don't have a server with clients I can instantly deploy and have it build all my existing jobs in a staging environment.
What I have, is a codebase that improves on shared commons code, that can deploy my jenkins master with options depending on what environment it is, and can deploy locally or onto my staging services. That's already a lot better than what I had before, even if it's not complete yet. For example, I know what pinned versions of what things work in combination currently, and what options matter to me.
I've found files that aren't checked in anywhere (that's a little scary), learned about the permissions set of the deployed environment vs. the current environment.
I've learned how to test the commons cookbooks, and how to ignore/extend them. And a short workflow for manually pumping my builds into my staging environment. And made plenty of mistakes (hard code private key and password values, users, etc). So there's plenty of data segmentation / securing ahead.
I've also learned a marginal amount of ruby, and started thinking more about machine state. And guarding from doing extra work, and started sending / subscribing to specific resource events.
I'm convinced I'm further ahead, even though I'm not complete yet. Because I have enough self-training now to move forward to the face-shooting part, where I try to dogfood and learn from deploy/rollbacks. Yuck. That won't be all that fun.
Best / worst of all, I have not been testing any portion of my work (besides iterating locally with chef-solo and in staging). That shit takes FOREVER. But, at some point, I will need to buckle down to ship common cookbook patches, and to trust my upgrade/downgrade/rollback paths.
Since I can sync the production server generated data onto a chef-managed system, I should be able to repeatedly test whether or not my upgrades will work in practice, and then, well, start working on features and new build chains in a managed fashion. That should be nice!
Leave a comment