#Gitlab cache
urban frigate
/cc @daring gazelle
Hey @graceful anvil you said use free gitlab runners, what do you mean?
How will that allow @cinder sparrow to add cache in CI
@cinder sparrow can you tell us what kind of artifacts you are looking to cache ?
Gitlab cache
daring gazelle
@cinder sparrow what's your current strategy to cache container builds in GitlabCI?
graceful anvil
Hey <@933501536624054272> you said use free gitlab runners, what do you mean?
...
I meant, free GitLab.com CI/CD runners or their own self-hosted runner on kubernetes, for instance. I'm assuming the former.
If the latter, they could employ strategies like we use for our CI in Kubernetes where we get node-level caching, etc.
This is an active area of research for us, trying to make this way easier for folks to DIY with commodity compute/storage. It's tough when you're using the "free" runners from GitHub Actions or GitLab.com sometimes, because they don't always have a great interface for the kind of caching that might help most.
That said, some folks do things like build a bunch of deps in a container image and push that to a registry, pulling it to do the builds with, etc. Have to look at the time it takes to pull certain things from some sort of colder storage vs having it in warmer cache. Some of the classic tradeoffs.
cinder sparrow
Hi guys 👋🏼,
can you tell us what kind of artifacts you are looking to cache ?
Hard question 😅. I have a really simple project where I run PHPStan/ECS/Unit tests, and I decided to just try the Dagger out.
So I have an access to default GitLab (or GitHub as I use both) artifacts mechanism.
My main issue is the first run is ~5 minutes locally, but every successive one is way faster. On CI every time I have to wait 5 minutes.
Currently I'm using GitLab hosted and GitHub hosted runners.
urban frigate
Hey @cinder sparrow
quartz compass
Personally, I stuck a dagger engine on my K8s cluster and it is pretty OK at keeping the cache around