chore: add new deploy job for new dogfood #10852
Conversation
|
Why do we need a new helm chart? We can just re deploy from the chart we already have on |
|
@matifali If we don't deploy the helm chart from main then we can't dogfood changes to the helm chart. It's not as important as the image obviously but we should still be running the latest helm chart to find bugs or try new helm features out |
|
If we deploy from the Something like: coder/.github/workflows/pr-deploy.yaml Lines 358 to 361 in 2c6e0f7 |
|
It won't look like that since it's all in Flux but I can try something along the same lines |
|
@deansheather OK. I am not experienced with Flux, so I am excited about it. Thank you for doing this :) |
That approach would also work, but having a separate Helm chart for coder-preview does make it easier to just install the 'latest' version of Coder from the Helm chart without having to worry about overriding the image. |
|
We can also move our wsproxies to fly.io VMs, then they can use the same coder-preview image, and we can simplify our logic considerably. |
We're not touching proxies for now, but if we're moving them we'd probably just move them to the dedicated servers that we already have in each of these regions |
.github/workflows/ci.yaml
Outdated
| flux --namespace flux-system reconcile source git flux-system | ||
| flux --namespace flux-system reconcile source git coder-main | ||
| flux --namespace flux-system reconcile kustomization flux-system | ||
| flux --namespace flux-system reconcile kustomization cert-manager | ||
| flux --namespace flux-system reconcile kustomization cert-manager-issuers | ||
| flux --namespace flux-system reconcile kustomization coder-certs | ||
| flux --namespace flux-system reconcile kustomization coder | ||
| flux --namespace flux-system reconcile source chart coder-coder |
There was a problem hiding this comment.
I could never find a way to tell Flux to just reconcile ALL THE THINGS :(
| "" = "tcp://dogfood-ts-cdr-dev.tailscale.svc.cluster.local:2375" | ||
| "us-pittsburgh" = "tcp://dogfood-ts-cdr-dev.tailscale.svc.cluster.local:2375" | ||
| "eu-helsinki" = "tcp://reinhard-hel-cdr-dev.tailscale.svc.cluster.local:2375" | ||
| "ap-sydney" = "tcp://hildegard-sydney-cdr-dev.tailscale.svc.cluster.local:2375" | ||
| "sa-saopaulo" = "tcp://oberstein-sao-cdr-dev.tailscale.svc.cluster.local:2375" |
deansheather
force-pushed
the
dean/dogfood-v2
branch
from
4e66630 to
f46721f
Compare
Changes the
build-main-imagejob to be justbuild,and now pushes helm packages tohttps://helm.coder.com/v2/previewon every push to main.One problem I foresee is that the index will grow very big very quickly but Helm doesn't have a good method of clearing old releases so I'm not sure what the best plan of attack is.I'm also unsure how to make Flux always pick the latest pushed version either without making github actions edit the manifests directly in the other repo. I'm guessing we need to mangle the version somehow to make it prefer the newest helm package.