Please provide example of multi-cluster container-native load balancing

[naseemkullah]

As per https://cloud.google.com/kubernetes-engine/docs/how-to/container-native-load-balancing ... thanks!

[morgante]

@naseemkullah As the linked doc explains, GKE load balancers can (and should) be created using Kubernetes objects.

Is there a specific ask you have not accommodated by the kubectl tutorial?

[naseemkullah]

Thanks @morgante, I am new to GCLB (used regional LB fronted nginx-ingress until now) and as my org is working on distributing services globally we want to use the GCLB.

Terraform is what we use to deploy GCP resources so I thought it would have to be created via TF.
My mistake!

[morgante]

No worries! You should be able to use the Kubernetes resources directly (the GKE controller handles creating the GCP resources for you).

[naseemkullah]

Great, thanks! That saved me a lot of messing around. :)

[naseemkullah]

@morgante My issue is I want a multi cluster ingress defined declaratively, is this possible without using Terraform?

kubernetes/ingress-gce#794

[morgante]

Got it, for that case you might want to look at autoneg: https://github.com/GoogleCloudPlatform/gke-autoneg-controller

[Dev25]

You could this today via standalone NEGs + this module with some tweaks (support max_*_per_endpoint backend group vars) + using the NEG datasource

I do have a large/breaking PR that includes NEG support #57 which includes an example of doing that.

You could make a smaller PR that just includes the minimum changes (those backend vars) to support NEGs

[naseemkullah]

Got it, for that case you might want to look at autoneg: https://github.com/GoogleCloudPlatform/gke-autoneg-controller

This looks really interesting @morgante

[naseemkullah]

You could this today via standalone NEGs + this module with some tweaks (support max_*_per_endpoint backend group vars) + using the NEG datasource

I do have a large/breaking PR that includes NEG support #57 which includes an example of doing that.

You could make a smaller PR that just includes the minimum changes (those backend vars) to support NEGs

That's excellent @Dev25

May I ask, once your PR goes through could you please confirm if this is the correct workflow:

1. Use TF to deploy GKE as well as this module to deploy the LB with backend services
2. Use whatever method to deploy your app to GKE, including the standalone NEG annotated service
3. Go back to TF and use it to configure this module, with the new standalone NEGs as datasource to be added as NEGS to the backend service created in step 1

[Dev25]

Yes that would be correct @naseemkullah , once you create the NEG service inside GKE you can figure out the compute engine NEG name (there is a cloud.google.com/neg-status annotation on the kubernetes service) and use that in TF/this module.

[naseemkullah]

Thanks for clarifying, I look forward to your PR being merged!