Has anyone had success with a single agent deployment handling multiple environments via labels? (more info in thread)

Hey @Leon Kozlowski, I’m not sure I am following 100%. Is there a use labelling dev and prod if they end up going to the same underlying compute? It will work though if your agent has labels

dev

and

prod

since flow labels just need to be a subset of the agent labels. Or is the main question how to assign

dev

to the flow if it’s in the

dev-xyz

project and

prod

automatically if it’s in the

prod-xyz

project?

I want to use a different set of secrets for dev and prod + the question is, is it an antipattern for them to exist on the same cluster

Ah ok do you have multiple tenants? I think it is an anti-pattern, but i can understand why it makes sense if you are changing all the connections and secrets.

No, do not have multiple tenants, the real problem is that were just using 1 cluster

Oh I see what you mean, and then you’d just set the environment variables different for those agents? Well, what if you have one cluster that has two

config.toml

files and you put secrets there, and then set

prefect.config.user_config_path

to point to the corresponding

config.toml

file for dev and prod. And then you can start 2 agents on the same machine with different env variables and secrets?

Acutally, could this just be done with namespaces?

I would imagine so, but not 100% sure because I’m not that good at kubernetes myself. But I see it working since you can have a different configuration for each namespace yep. Then you would have one agent per namespace right?

that would be ideal, but its somewhat out of scope for our helm workflow that were using to deploy the agent - I’ll see if I can make that happen - thanks @Kevin Kho

Glad you got it working 👍