State of the Art Deployments

I’ve been thinking about deployment automation lately, especially because things at work are getting hectic. We’ve got serveral sites and several geographies to serve. We also engage with a handful of firms to do design work and some development work. While this is a totally reasonable approach because we aren’t primarily a technology company, it does mean that there are a lot of fingers in a lot of pies.

I think that we are going to need a way to deploy branches of the project on to ephemeral servers (preemptible vms/spot vms/low priority vms depending on who your favored cloud provider is). I also think that we are going to need to dockerize these applications to unify the differing development, testing and deployment environments.

Just like everyone else in the tech community, super complex systems with YAML files and dashboards are like a siren - incredibly enticing - but never the less dangerous.

In the past git based deploys have been the name of the game, capistrano for running rake files and Bamboo because of some organizational requirements.

This blog by the way is deployed by a bash script called deploy.sh. It contains a single line. Just throwing some files in a directory served by apache is still a great strategy.

jekyll build && scp -r _site/* root@chuckgreenman.com:/var/www/html

Brevity is the soul of wit or in this case, the soul of not using kubernetes to deploy your dinky personal blog.

- @chuck_greenman