Just checked quickly, with the Pod resource: setting the provider like above does work

https://github.com/pulumi/pulumi-kubernetes/blob/dbc3697651b01ba5de3859a0dcf66003d45d8319/pkg/gen/node-templates/provider.ts.mustache#L127 ConfigGroup looks like it should take ComponentResourceOptions

Oh, that was changed with #140, 5 days ago ...

In my local version:

class ConfigGroup extends CollectionComponentResource { constructor(name: string, config: ConfigGroupOpts, opts?: pulumi.CustomResourceOptions); }

You could probably just force a cast

Although this makes me think of two things: A) The error message could be better to tell users to look at ComponentResourceOptions.providers B) ComponentResources that only work with one provider type should probably just take CustomResourceOptions and translate CustomResourceOptions.provider to ComponentResourceOptions.providers keyed by the one provider they're using.

If I use pulumi/kubernetes 0.15.2-dev-1534791785-g877633f, then it works as @helpful-vegetable-35581 suggested (providers: {kubernetes: k8sProvider})

Thanks for the help!

n/p 🙂

Sorry or the troubles here. If you are using the latest released versions, then you should be able to pass the provider via to a component via:

providers: {kubernetes: k8sProvider}

. We've seen a few folks trip up on the difference between how CustomResources and ComponentResources get parameterized by providers, we're considering some sugar to make this simpler: https://github.com/pulumi/pulumi/issues/1802. This is effectively the same idea @helpful-vegetable-35581 notes in (b) above. cc @microscopic-florist-22719

@white-balloon-205 I was using 0.15.1 when the error happened, only the 0.15.2-dev-1534791785-g877633f fixed it for me (pulumi/kubernetes)

I think I found another bug / inconsistency: in kubernetes "default" namespace can be omitted from manifests. But if I read the following manifest file: apiVersion: apps/v1 kind: Deployment metadata: name: bzcerts-cert-manager namespace: "default" ... Pulumi creates the object (internally) with name: default/bzcerts-cert-manager However if I use this: apiVersion: apps/v1 kind: Deployment metadata: name: bzcerts-cert-manager ... Pulumi creates the object (internally) with name: bzcerts-cert-manager It tripped me up when I try to get some info from the spec: export const certManager = configGroup.getResource("apps/v1/Deployment", "bzcerts-cert-manager"); It works only with the second manifest format, but it should work the same for both, I think. Should I file a bug for this?

Is there any way to "force" destroy a stack? I got a CRD deployed to my Kubernetes cluster, but destroying it fails, and it stops the whole destroy process

(btw: the CRD is still present in the cluster, but somehow Pulumi doesn't recognize it)

@gray-city-50684 @helpful-vegetable-35581 Thanks for the report, we’ll cut a v0.15.2 today with fixes for the first and third problems + the Helm stuff from yesterday.

I’ll respond in-thread for the second problem, about namespace naming.

@creamy-potato-29402 Sounds great, thanks!

@gray-city-50684 ok, I’ve filed the following to track #2 and #3 of the issues you bring up, but on further inspection (as you mention), #1 is actually already fixed https://github.com/pulumi/pulumi-kubernetes/issues/146 https://github.com/pulumi/pulumi-kubernetes/issues/147

Nice, I think 145 could be then closed, pointing to 147 (I made 145 before talking to you about the issue)

ah shoot

my bad, sorry

New day, new way to fail

pulumi refresh

🙄

i just go to starbucks and back 😕

I'm trying to create an Azure DNS entry based on the external IPs of a k8s.core.v1.Service object (nginx ingress controller created from YAMLs)

if (config.dnsUpdateEnabled) { const dnsEntry = new azure.dns.ARecord("mydns", { name:

${pulumi.getStack()}.myservice

, resourceGroupName: config.dnsResourceGroup, zoneName: config.dnsZoneName, records: nginxControllerService.spec.apply(spec => spec.externalIPs), ttl: 3600 } ); }

Diagnostics: global: global error: azure:dns/aRecord:ARecord resource 'bzdns' has a problem: "records": required field is not set

What is the correct way to "hook up" an output value from an object like that, as the input to another?

is the relationship between vendor bills and the custom record many to one? i.e. multiple bills select the same custom record