Hacker News new | ask | show | jobs
Using a Docker registry as a distributed layer cache for CI (blog.meadsteve.dev)
23 points by meadsteve 1353 days ago
5 comments

We do this for most of our docker image builds in CI. Definitely worth it if your CI builds take significantly longer than it takes to pull down the image layers from previous builds (sometimes network speeds can end up slowing it down with a method like this)
We use it a lot. We also had a lot of issues with an official registry. Now we use https://goharbor.io/
Harbor is amazing - use it a ton for imgs and charts. It just works, is lightweight, and airgap friendly.
buildkit has a way to cache images in the GitHub Actions cache, making this full circle: https://github.com/moby/buildkit/blob/47e953b2/README.md#git...
This article is weirdly vague in the sense they didn't mention the registry is is an official docker container: https://hub.docker.com/_/registry

It also doesn't go into any of the detail about starting up / configuring the registry, or explaining that running your own version is like running you own private hub.docker.com.

The registry is great! It just doesn't seem to be maintained much anymore, maybe because it's feature complete, or maybe because Docker is trying to sell plans to their "hub".

For anyone else interested in advanced Docker features: check out Docker Swarm (the only way to reasonably run Docker Swarm is to manage your own Docker Registry).

Well I already had a docker registry. I generally just blog the information my past self would have needed a few weeks ago.
I already have a docker registry too, so I guess I missed the point of your entire article. You can use the registry as a cache for build artifacts... that's what it's for. Your article was written like you just figured that out, and I was providing context to the other confused HN users.
Somehow I'd completely missed this feature existed and it's a really nice way of speeding a build up