[Beats] Custom configmap for Kubernetes Orchestrator Name

[mholttech]

Describe the enhancement:
#26056 introduced 3 methods to set orchestrator.cluster.name & orchestrator.cluster.url

1. kube_config file
2. kubeadm-config configmap
3. GKE Metadata

I would like to see a separate configmap for deployments that aren't managed by kubeadm or GKE. While one can create a simple kubeadm-config configmap, this may cause other services that key off of the existence of this configmap to not work as expected.

Describe a specific use case for the enhancement or feature:
I have an EKS Cluster with Managed Node Groups and utilize Fleet containers in Kubernetes for monitoring. I was able to get the cluster name to populate by creating the kubeadm-config config map, however when I went to deploy Calico it saw this configmap and it was missing required information, causing Calico to not deploy successfully.

This is the configmap I deployed:

apiVersion: v1
kind: ConfigMap
metadata:
  name: kubeadm-config
  namespace: kube-system
data:
  ClusterConfiguration: |
    clusterName: cluster-1
    controlPlaneEndpoint: https://abcd1234.gr7.us-west-2.eks.amazonaws.com

[mholttech]

@bvader Can you please make sure this gets routed to the appropriate team? Thanks!

[tetianakravchenko]

Hi @mholttech!

Orchestrator fields are not supposed to be defined "manually" by custom configMap. The proper way here would be to use something similar to the GKE Metadata, but for EKS. I checked it quickly and seems there is nothing similar in EKS now - aws/containers-roadmap#686.

Meanwhile I would suggest to use rather add_fields processor and add orchestrator.cluster fields explicitly, instead of using kubeadm-config to avoid unexpected effects:

processors:
  - add_fields:
      target: orchestrator.cluster
      fields:
        name: cluster-1
        url: https://abcd1234.gr7.us-west-2.eks.amazonaws.com

As EKS metadata is missing, I'm going to adjust documentation with add_fields processor and close this issue.

[ottramst]

Hi!

Just clarification on this topic - so this means that when running Elastic Agent in EKS, there is currently no way of adding the orchestrator.cluster.name in the Agent configuration?
Seems a bit much to have an ingest pipeline just for adding a single field.

[tetianakravchenko]

Hi @ottramst

you don't have to use ingest pipeline, you can add add_fields processor in the agent configuration, example - https://github.com/elastic/elastic-agent/blob/main/deploy/kubernetes/elastic-agent-standalone-kubernetes.yaml#L291-L295

[vnirocha]

Hey @tetianakravchenko

I'm running fleet managed elastic agents and there is no way to use processors on kubernetes integration. It seems like the team is already working on this feature (elastic/integrations#4363) but, until this moment, there is no way of doing, unless using the standalone implementation of the elastic agent.

[tetianakravchenko]

Hi @vnirocha

I've just asked to promote the change you linked - elastic/integrations#4363 (comment). I believe version 1.26.0 should be available in next few days.
Please note that this change only applies to the elastic stack versions >= 8.4 https://github.com/elastic/integrations/pull/4363/files#diff-d20d8086b58468701d8a944fc3ead31ddcc098054affdcc335b69464b4c16c54R13

[radhane]

Instead of adding processors which do not work in my case for all metrics in elastic 7.x with managed kubernetes, I have used a separate namespace with cluster name instead of adding all clusters in the default namespace. And used data_stream.namespace as cluster name filter.