the issue is the stuff i want to change isn’t exposed by k8s

(i want to have the LB just proxy traffic to another service, basically SNI-based https->http for a help site that is hosted somewhere else)

I am about to just push up a tiny nginx container that does that proxying and point the backend at that, but its not optimal considering GLB can do it directly.

Let me see if I understand the situation. You have an ingress controller on a K8s cluster that is exposed via a Service of type LoadBalancer (and thus creates a GCP LB). You want the ingress controller to direct traffic to a site that is external to the Kubernetes cluster, is that correct?

no, its an ingress, not a loadbalancer

which due to me using GCP causes it to be backed by GLB

And yes, I want it to proxy traffic from

help.domain

to

some.other.domain

OK. Which ingress controller are you using? Not all of them support what you're trying to do.

GLBC

Gotcha. Reviewing the GLBC docs, it's unclear if it supports your use case. I haven't used it personally, so I can't say. My initial guess would be that it probably won't support your use case, and you're probably better off just using a GCP LB directly instead of trying to manage it through Kubernetes ingress objects.

I think the Nginx controller supports this use case, if memory serves correctly.

the cert is managed inside k8s

so I kinda have to go in via the main GCP LB

I think I am going to just run nginx inside a container and have it proxy for now

get deeper later on

Sure

thats how i do that on azure - a combination of cert-manager helm and nginx-ingress running inside the cluster

works fine with the azure loadbalancer since we never have to interact with it directly

I see the ability to tag ecs services was added in this pr: https://github.com/pulumi/pulumi-awsx/pull/179/files. It does not seem to include

tags

for

awsx.ecs.FargateService

. Was there a reason this was omitted?

@better-rainbow-14549 yeah, figured as much

@salmon-account-74572 thanks!

It seems like switching to the ingress-nginx may be a good idea in the near term anyway!

I’m trying to write a run time test based on this blog post. I copied the bucket test from the Runtime Testing, but it fails to build. Am I doing something wrong, or is this just a bad dependency in the package? https://blog.pulumi.com/testing-your-infrastructure-as-code-with-pulumi

❯ go test -v
go: finding <http://github.com/pulumi/pulumi/pkg/testing/integration|github.com/pulumi/pulumi/pkg/testing/integration> latest
go: finding <http://github.com/pulumi/pulumi/pkg/testing|github.com/pulumi/pulumi/pkg/testing> latest
go: finding <http://github.com/pulumi/pulumi/pkg|github.com/pulumi/pulumi/pkg> latest
go: finding <http://github.com/Nvveen/Gotty|github.com/Nvveen/Gotty> v0.0.0-20170406111628-a8b993ba6abd
go: <http://github.com/Nvveen/Gotty@v0.0.0-20170406111628-a8b993ba6abd|github.com/Nvveen/Gotty@v0.0.0-20170406111628-a8b993ba6abd>: unknown revision a8b993ba6abd
go: error loading module requirements

@freezing-flag-53289 I had that too this weekend. You can solve that by adding

replace <http://github.com/Nvveen/Gotty|github.com/Nvveen/Gotty> => <http://github.com/ijc25/Gotty|github.com/ijc25/Gotty> v0.0.0-20170406111628-a8b993ba6abd

to your

go.mod

file

that solved the issue for me 🙂

I'm seeing some unexpected behaviour managing s3 events with

bucket.onObjectCreated

and

bucket.onObjectRemoved

. If I have multiple events defined (which is perfectly valid in s3), removing 1 removes them ALL. Is this a known bug?

I’m getting the following error:

The Pulumi runtime detected that 114 promises were still active
    at the time that the process exited. There are a few ways that this can occur:
      * Not using `await` or `.then` on a Promise returned from a Pulumi API
      * Introducing a cyclic dependency between two Pulumi Resources
      * A bug in the Pulumi Runtime

It definitely is the second case: a cyclic dependency, but I’m not sure how to break the cycle. I’ve got a lambda that closes over a function that constructs a config/context object sort of thing, but another lambda that needs the arn of that first lambda - is there some general trick to break the cycle in this sort of situation?

big panic trying to add config to GCS buckets - will file a bug if it looks appropriate: