that might help

(

-d

enables debug logging)

all 180 lines? 🙂

well, do you see an exception in there? 😄

hey, can you validate my approach? I've created a pulumi configuration that creates couple of resource in Azure and some deployments\service in k8s. I want to run that configuration in a pipeline for each branch developers build (so they can test their stuff). Am I correct in assumption that I need pulumi stack init for each deployment so they dont collide? also, how do I delete pulumi resources and stack afterwards (assuming build agents got no state). also, how do I auth silently to pulumi state server (I wonder if I even need that, since I dont really manage state, I just create stuff and at some point in time I need to clean it up).

We have some docs for that very question: https://pulumi.io/reference/cd.html

For your stack question: https://pulumi.io/reference/stack.html

the short answer: yeah, one stack is roughly equal to one deployment.

I saw the stack article, it doesnt really clarify a lot to me

ok, that (your short answer) helps

PULUMI_ACCESS_TOKEN - I can just use build variable here, right?

yep - an environment variable in your build config.

thanks, going to do some experimenting

do you guys have official docker images?

we do, yeah: https://hub.docker.com/r/pulumi/pulumi

does that contain all node\python sdk?

no, it is the CLI only. Your language’s package manager is responsible for installing the language SDKs.

mhm, but does it contain python?

Yes - it does include Python. See https://github.com/pulumi/pulumi/blob/6f386567f6a8494853bee8fccdb455b3be1a993b/dist/docker/Dockerfile#L28.

ok, just that page doesnt link to dockerfile

Opened https://github.com/pulumi/pulumi/issues/2432.

ok, just so you know I'm not wasting your time. I'm certainly using this with at least one client, going to blog about this and probably throw some examples in the example repo. and i'll prolly talk on at least one meetup about this 😉

i really like the python stuff, i think that's a move in the right direction. only thing that's really lacking right now is debug experience, which is more or less non existent 😞

FWIW - debugging support is tracked in https://github.com/pulumi/pulumi/issues/1372. Definitely want to add this!

yeah, Sean mentioned that

hey folks, another stupid question, how can I validate my k8s "manifests"? the thing is, they are using a slightly different notation sometimes. questionmark. yaml has selector. matchLabels, but pulumi has selector.match_labels

https://github.com/pulumi/pulumi-kubernetes/blob/master/sdk/python/pulumi_kubernetes/apps/v1/Deployment.py no validation here, not sure where do I look 🙂

another question: i pass the default namespace to the k8s provider, but all the resources are being created in the default ns, is this expected?

mhm, any chance there is a bug? I dont see anything wrong with this definition:

RoleBinding(
    "orleans-rbac-rb",
    metadata={
        "name": "orleans-rbac",
        "namespace": "default"
    },
    role_ref={
        "apiGroup": "<http://rbac.authorization.k8s.io|rbac.authorization.k8s.io>",
        "kind": "ClusterRole",
        "name": "orleans-rbac"
    },
    subjects=[
        {
            "kind": "ServiceAccount",
            "name": "orleans-rbac",
            "namespace": "default"
        }
    ]
)

this binds to a clusterrole, but thats intentional. it should be the reason to fail anyway. no way pulumi can know I have or not have that role, nor should it care

same thing works in yaml