[BUG] Node pool and Dashboard pod affinity settings are not taken into account when deploying via Helm

[dmantas]

What is the bug?

When deploying a cluster via Helm chart, defining affinity under dashboards or under one of the nodePools is not taken into account. Checking the Dashboards Deployment or the Node Pool StatefulSet shows that the pods do not have any affinity settings at all.

If the cluster is deployed by using kubectl apply on the OpenSearchCluster manifest, affinity is applied to the pods.

How can one reproduce the bug?

Create a values.yaml file where under nodePools: the following exists:

    - component: data-nodes
      diskSize: "5Gi"
      replicas: 3
      roles:
        - "data"
      resources:
        requests:
          memory: "2Gi"
          cpu: "500m"
        limits:
          memory: "2Gi"
          cpu: "500m"
    - component: cluster-managers
      diskSize: "1Gi"
      replicas: 3
      roles:
        - "cluster_manager"
      resources:
        limits:
          memory: "2Gi"
          cpu: "500m"
      affinity:
        podAntiAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
            - labelSelector:
                matchExpressions:
                  - key: "opster.io/opensearch-nodepool"
                    operator: In
                    values:
                      - "cluster-managers"
              topologyKey: "kubernetes.io/hostname"

What is the expected behavior?

When checking the cluster-managers StatefulSet definition:

 k get statefulset.apps/caas-opensearch-cluster-managers -o yaml   

we expect to see the affinity settings:

    spec:
      affinity:
        podAntiAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
          - labelSelector:
              matchExpressions:
              - key: opster.io/opensearch-nodepool
                operator: In
                values:
                - cluster-managers
            topologyKey: kubernetes.io/hostname

However, there is no affinity key at all under spec.

What is your host/environment?

Opensearch k8s operator version: 2.4.0
Opensearch version: 2.8.0

[evheniyt]

I have also prepared PR #640 with a fully refactored helm chart that solves such kind of issues (not only affinity but in general issues with missing templated configuration)

[prudhvigodithi]

[Triage]
Thanks @evheniyt I see the PR is closed, can you please open again? AFAIK there should be a bug in https://github.com/opensearch-project/opensearch-k8s-operator/blob/main/charts/opensearch-cluster/templates/opensearch-cluster-cr.yaml.
Adding @swoehrl-mw @jochenkressin @pchmielnik @salyh @bbarani

[evheniyt]

@prudhvigodithi it was closed by my mistake, now I have reopened it.