Back Up And Running

I’m back. The last couple of months have been hectic, from a personal standpoint. I walked away from my rancher instance, and dealt with what I had to deal with.

HELLO WORLD!

When I was able to dedicate some time back to engineering in my free time, this is what I found:

  • One root hardrive failed, taking the whole server with it (no big deal).
  • When the server failed, rancher moved gitlab to another server, updating it in the process (why :latest is a bad idea)…
  • gitlab fell over on itself, and I don’t have the resources — aka, time — to bring it back.

So, what do you do? Why, you reverse engineer the work you spent most of a couple years working on, questioning everything.

  • Was it really necessary to use a CI to build things?
  • Was it really necessary to use memcached/rethinkdb?
  • Can I cut out this part of the pipeline and make it faster/simpler?

Along the way, at work, we needed a more robust/simpler build system as did I for my blog. So, I invented (borrowing heavily from make and gulp and npm) this interesting little tool called tyche aka, cutter3 (cutter 1-2 is closed source and tightly coupled to the project it builds)

Tyche

Tyche allows me to declaratively define how to build things in various architectures. I can explain to tyche how osx should build, vs linux or windows. Effectively allowing me to declare development vs production, which would be a nice feature to have built-in in a later version.

Currently, I’m designing a declarative language for it. Designing a language is the hard part, the rest is easy. I think I’m currently on my fourth iteration of the language.

I think this is a great way to get a project to build itself, similar to make, but deliberately simpler.

But, shhh, I think make is better!

So, what happened to the blog?

I’m currently running a cluster in Azure, and focusing on the stateless, no-shared aspect. Everything is built locally, pushed to Quay.io, and deployed to rancher, all using tyche.

The beauty of Docker, is that any developer can build the same thing, across any architecture. So, why have a dedicated server do that? Exactly. We don’t need one.

The future…

Now that I am able to blog again, I plan on doing just that…