Now I just have to find a Windows machine to test on :)

I had to give mine back when I left MSFT

It shouldn't be too hard, I think we can put a fix together today...

If you have a fix in mind, I can test it locally (not a big TS coder, but maybe this much I can manage 🙂 )

I think Matt or Cyrus has a Windows machine, I want to make sure there aren't other issues too

Should be pretty quick to fix I've they're in

Hi! I've been thinking of a scenario when a piece of cloud stuff is provisioned per tenant on signup... Can pulumi help in such use case? Seems like CLI doesn't really fit then.

A stack per tenant? pulumi stack new <tenant>; pulumi update; ?

Indeed - this is possible today by invoking

pulumi

CLI commands on this user signup event, and managing these as separate Pulumi stacks.

Or even one stack, and make part of the config a list of all tenants and just wrap the whole program in a loop through that

We have actually considered a Pulumi Pulumi provider, which would allow managing the deployment of Pulumi stacks using another Pulumi program, which might provide a slightly more first-class approach to this.

All of these are for fairly coarse-grained multi-tenancy. For finer grained multitenancy, it may make sense not to provision the resources using Pulumi, and instead to manage the granular tenants using code at runtime.

what is a granular tenant?

I just mean it depends how granularly you are alloacting resources to tenants - like will you share a database across tenants and isolate via some key per tenant?

I might want a mix like a global Azure SQL elastic pool + a DB per tenant

Sharing a database is old and well-understood but there is an opinion that it's time to leave this job to cloud providers

I'm just flirting with this idea in my head

Probably not that interesting given that the Kubernetes provider for Pulumi already does this, but I just ran into this which is interesting: https://github.com/dikhan/terraform-provider-openapi

How often do you update terraform providers? I'm interested in the recent fixes of the azurerm provider (https://github.com/terraform-providers/terraform-provider-azurerm/pull/1769).

I upgraded to 0.15 + all package versions too, and now when I run

pulumi refresh

I get

error: could not find plugin for provider 'urn:pulumi:myproject::myproject::pulumi:providers:aws::default'

. What have I broken?

Because of the bug I found yesterday with using Helm charts with Pulumi on Windows 10, I thought I'll turn those charts into Kubernetes manifests with "helm template ..." and use them directly. I make my K8S cluster with pulumi/azure like this (and this part works great): export const k8sCluster = new azure.containerservice.KubernetesCluster("MyCluster", { ... }); export const k8sProvider = new k8s.Provider("MyCluster-Provider", { kubeconfig: k8sCluster.kubeConfigRaw, }); And then I try to apply the manifest yaml's like this: const certManagerCG = new k8s.yaml.ConfigGroup("cert-manager", { files: "yaml/cert-manager/*.yaml" }, { provider: k8sProvider } ); And I get the error: Diagnostics: pulumi😛ulumi:Stack: beyondzen-dev info: error: [runtime] Explicit providers may not be used with component resources

Could somebody please tell me, how to do it correctly?

I think you want providers: {"k8s": k8sProvider}

See "providers" on the ComponentResourceOptions interface

I tried that, but ConfigGroup wants CustomResourceOptions

and in the example for using Pods, it was like this:

const nginxcontainer = new k8s.core.v1.Pod("nginx", { spec: { containers: [{ image: "nginx:1.7.9", name: "nginx", ports: [{ containerPort: 80 }], }], }, }, { provider: myk8s });