https://pulumi.com logo
Join Slack
Powered by
This message was deleted.
# kubernetes
s
This message was deleted.
b
b
Tried it but does not work, transformations seem to work for other labels but 
managed-by: pulumi
seems to be specially treated by pulumi
wait, I tried transformations inside 
ConfigFile
docs, but haven't tried them on resource level, let me check
m
@better-dentist-3207: are you using 
renderYamlToDirectory
to generate the yaml files?
b
yes
If I set this up right I'm planning to write an ArgoCD Plugin to just use the generated files with specific directory patterns through ArgoCD
m
then activate SSA too in the provider and you should not get the managed-by
b
Do those two options not conflict with each other? Asking because this pulumi project needs to work fully offline, even if a valid 
.kube/config
is detected.
m
i did a similar thing for a workshop with Flux, but rendered to a S3 bucket to pick up: civo-navigate-gitops/gitops/baseconfig.ts
b
That's a great repo, thanks!
This structure seems really similar to what I'm currently doing. Abstract 
pulumi.ComponentResource
as something like 
ArgoCdApplication
with a built in k8s provider, and have people extend that. Then all 
ArgoCdApplication
instances get passed to a 
Bootstrap
ComponentResource instance which creates actual ArgoCD Custom Resources
One issue I haven't solved yet is because everything is in a single pulumi project, the argocd plugin has to render every app for each time it needs to extract a single app.
m
Split the Pulumi project into separate ones?
b
Can do that, but I'm trying to keep this solution as stateless as possible for just generating manifests. Generated ArgoCD 
kind: Application
Custom resources are dependent on the names of the ComponentResources. If I start introducing state with stack references, things start getting janky on the side of the ArgoCD Plugin
m
valid point!
10 Views
Open in Slack
PreviousNext
Powered by