Docker Layer Cacheing optimization: Docker layer affinity for volumes
R
Ryan Vangundy
I'm currently doing a lot of work that builds a Docker image using Docker Layer Cacheing, but I'm doing so in a branch. At the same time, several other developers are also working in the same project but the Dockerfile for this image isn't yet merged. As a result, their Docker remote engines are acquiring volumes so starve my branch of cached images. This makes working in a branch really slow because it seldom chances on a volume that has the image layers available. If jobs had some kind of affinity for mounting volumes that had available image layers, this would improve branched development a ton!
CCI-I-800
M
Matthew Morrissette Vance
Absolutely agree. The docker layer caching should work similarly to the other save_cache/restore_cache where there are "keys" and it tries to find the machine which has the matching key.
D
Derek VanHove
I'd go a bit further with this and propose some sort of cache tagging, allowing developers to target a bit better different cache volumes for specific jobs. For example, we build multiple docker images in the scope of a single project through a generic job pointed by a child job. It would be useful to be able to dictate a tagged cache to mount to guarantee that the right cache volume is available