[FEATURE] Allow namespaces to be over-ridden in all addons (prometheus and others)

[spkane]

Please describe your quesiton here

The namespace for the prometheus addon is hardcoded. Why is this?

I have spent a bit of time trying to change this, before realizing why I couldn't get it to change as expected.

Provide link to the example related to the question

terraform-aws-eks-blueprints/modules/kubernetes-addons/prometheus/locals.tf

Line 8 in deec7d5

namespace = local.name

Additional context

Similar issue/fix: #287

More

• [X ] Yes, I have checked the repo for existing issues before raising this question

[bryantbiggs]

I can't answer the why but I think instead we should default to what the associated addon's helm chart provides while offering the users the ability to change to suite. Changing this from a question to an enhancement

[spkane]

Agreed. I just wasn't sure if there was a reason for this, so I figured I'd start by asking.

[vara-bonthu]

@spkane What you are seeing in locals.tf are the default values for each Add-on. All the Add-ons can be configured by the users to pass their own values.

Did you try passing your own config with prometheus_helm_config={} ?

Here is the doc https://aws-ia.github.io/terraform-aws-eks-blueprints/add-ons/prometheus/

[spkane]

@vara-bonthu I had this setup in my code when I was testing it:

  enable_prometheus                   = true
  prometheus_helm_config = {
    version          = "15.8.5"
    create_namespace = true
    namespace        = "monitoring"
    values           = [templatefile("${path.module}/helm_values/prometheus-values.yaml.tftpl", {nodeSelector = local.primaryNodeSelector})]
  }

but since it has been a week, let me test it one more time and double check.

The values template looks like this:

# See:
# https://github.com/prometheus-community/helm-charts/blob/main/charts/prometheus/values.yaml
#
# This is a template, so we can get values from terraform.
#

alertmanager:
  nodeSelector: ${jsonencode({ for ns_key, ns_value in nodeSelector : ns_key => ns_value })}

server:
  nodeSelector: ${jsonencode({ for ns_key, ns_value in nodeSelector : ns_key => ns_value })}
  retention: 1h

  resources:
    requests:
      cpu: 500m
      memory: 512Mi

  global:
    scrape_interval: 15s

pushgateway:
  nodeSelector: ${jsonencode({ for ns_key, ns_value in nodeSelector : ns_key => ns_value })}

[spkane]

So, it looks like the monitoring namespace is created, but the ArgoCD application/helm is not targeting the right namespace.

Digging in just a bit more...

[spkane]

Ahhh. I think the error might actually be here:

https://github.com/aws-samples/eks-blueprints-add-ons/blob/main/chart/templates/prometheus.yaml#L33

It looks like the destination namespace is hard-coded here (and maybe in other examples) and does not respect what was set in the Terraform code.

[vara-bonthu]

Ahhh. I think the error might actually be here:

https://github.com/aws-samples/eks-blueprints-add-ons/blob/main/chart/templates/prometheus.yaml#L33

It looks like the destination namespace is hard-coded here (and maybe in other examples) and does not respect what was set in the Terraform code.

Yes, you are spot on here. Fix needs to go in this repo and the eks-blueprints-add-ons to pass the namespace to all the Argo enabled add-ons

[bryantbiggs]

Resolved here with #595

Similar issue exists for addons repo aws-samples/eks-blueprints-add-ons#43