Is there a way for me to generate YAML out of Pulumi kubernetes object without actually creating that specific resource into the cluster? I know this is a very specific request but the reason I would need this is that a 3rd party app ran on our k8s cluster supports providing a job template yaml to it as a file (yes, as a file, it doesn't support ConfigMaps etc). I would like to generate that file based on other cluster data that Pulumi is controlling. Generating a valid YAML file out of Pulumi resources with

Output

abstraction is not a trivial task and I would assume Pulumi is actually under the hood generating YAML anyway from the definitions it would make sense if one could also export that data out somehow. However, I think this is not possible since as far as I can see it, Pulumi automatically registers a resource to state when you say

new <resourcetype

and you can't override this functionality, right? If the answer to my question is what I think it is then my follow up question is: is there a ready tool or library function that could take in TypeScript data (objects, arrays etc) with Pulumi

Output

and it would return a valid YAML string with

Output

values properly populated? If the answer to that one is "no" then I have one additional question: how can I reliably generate a string dynamically recursively in such a way that

Output

objects are handled correctly? I actually wrote this simple "data to YAML" converter where I tried to handle the

Output

objects correctly but unfortunately it doesn't seem to work that well...

yes, use

renderYamlToDirectory

on the provider: https://www.pulumi.com/docs/reference/pkg/kubernetes/provider/#renderyamltodirectory_go

Cool! So, how does it actually work? Doesn't look like self-explanatory to me. "to directory" - so what does it do? Render it to a yaml-file, not to a a string?

You can construct a JavaScript object representing the structure you want, and serialise it once those outputs are known.

@stocky-spoon-28903 yes, definitely. But how? I mean, it's trivial to generate YAML, but with

Output

objects involved it isn't anymore. Let's assume I have a longish data structure with tens of Outputs. The only way I can think of that this could be done is to list all of the Outputs in

pulumi.all

and then use that to apply the values at the same time and then return the result. To me this feels messy and hard to read. I would much prefer that I could just place the Output values directly into the data structure where they belong to and call a single function to evaluate it in such a way that all Outputs within it are realized. Then I could just serialize that to YAML without issues

Hmm. I did try that obviously and that did not work