Feature #47286
cephadm: Local registry setup
0%
Description
The idea is to optimize cluster deployment time by creating a local registry that will then be used to pull images from other nodes.
One idea is to use the new service_type: container
generic deployment to deploy a registry.
Problem is: we need a ceph cluster in order to deploy the registry.
Therefore:
- Either we call `cephadm deploy ...` in `cephadm bootstrap` manually before starting the first MON,
- or we run bootstrap and modify
container_image
afterwards to point to the correct registry.
Note, we have to modify the registries.conf on all hosts for this to work.
Maybe this Plus a systemd unit:
podman run --privileged -d --name registry \ -p <option>REG_HOST_PORT</option>:5000 -v /var/lib/registry:/var/lib/registry \ --restart=always registry:2
History
#1 Updated by Sebastian Wagner about 3 years ago
- Tags set to low-hanging-fruit
#2 Updated by Sebastian Wagner about 3 years ago
I don't think this requires any code changes. this issue is about figuring out how to call the cephadm binary to make it deploy a local registry
#3 Updated by Sebastian Wagner almost 3 years ago
- Description updated (diff)
#4 Updated by Tim Serong almost 3 years ago
I'm not sure I like this idea. I mean, I know it's a useful optimisation to have a local container registry in general, and I know that air-gapped environments are a thing, where a local registry is necessary due to security policy, but do we really want to have the container registry be part of, and managed by, the cluster itself?
#5 Updated by Sebastian Wagner over 2 years ago
- Status changed from New to Fix Under Review
- Pull request ID set to 44092
#6 Updated by Sebastian Wagner over 2 years ago
- Status changed from Fix Under Review to Resolved