It's nothing fancy - I just use a typical k8s Service/Deployment object on GKE. A deploy is:

  1. docker build (most layers cached) - 2s
  2. docker push - 2s
  3. update deploy.yaml
  4. kubectl apply -f deploy.yaml
  5. kubectl rollout status deployments {name} - 6s

Lucky!! With our infrastructure, deploys can take an hour or so. 10 minutes for the build, 10 minutes for the image to get built, plus the rest of the time for terraform to apply infrastructure changes across dev, staging, and prod. Only thing we do have is automated testing after every deploy so issues tend to get caught. But that's still so long for a deploy! I don't know of a good way to get it down faster.

Why is terraform deploying infrastructure for every container deployment? Can't you just rollout onto the existing infrastructure?

Also sounds like there is some lay hanging fruit available by adding some caching/layering in the build process

Ah, I should have explained more. That's an hour for our ASG + EC2 deployments. The only benefit we get is easy roll backs because it always deploys a new ASG. We're switching over to Spinnaker and started with our EC2 infrastructure. I think container deployments will be faster but still, an hour for EC2 deployments!

Yeah this is exactly what I do too, works just fine. You probably already have something like this, but I hacked a bash+yq script that automatically updates all relevant yaml files with the latest image tag. So getting new code running is two lines:

make image push deploy

kubectl -f somewhere/deploy.yaml

Take a look at skaffold. Ours is under 10s