Enable ignoreResourceUpdate by default with known high churn resources

[agaudreault]

Summary

The feature is behind a flag. We should enable resource.ignoreResourceUpdatesEnabled: true by default the argocd-cm ConfigMap.

Motivation

No undesired effect has been reported since it has been implemented. Argo operators will likely experience problems until they find that flag and turn it on. If it is on by default, argo performance will be sustainable with default configurations for longer. CPU will be less and new/existing users are less likely to deploy Argo on instances with a lot of CPU request/limits, saving cost.

Proposal

• Set it to true by default
• For widely used Kubernetes resources that are known to have high churn, we should configure the ignore by default.

[andrii-korotkov-verkada]

Can we enable ignore updates and differences for the status field for all resources? Or is it for v3?

[rumstead]

FYI - Since v2.14 ignoreResourceUpdate is enabled by default!

Here are some suggestions on what to ignore from the docs

I don't quite follow the first one for ownerReferences.

  resource.customizations.ignoreResourceUpdates.argoproj.io_Application: |
    jsonPointers:
    # Ignore when ownerReferences change, for example, when a parent ApplicationSet changes often.
    - /metadata/ownerReferences
    # Ignore reconciledAt, since by itself it doesn't indicate any important change
    - /status/reconciledAt
    # status can change when controllers update a resource since it is used to track the status 
    - /status

Maybe we don't need /status if #10312 is merged

Some ones I use:

    jqPathExpressions:
    # Ignore lastTransitionTime for conditions; helpful when SharedResourceWarnings are being regularly updated but not
    # actually changing in content
    - .status?.conditions[]?.lastTransitionTime
  resource.customizations.ignoreResourceUpdates.all: |
    jqPathExpressions:
    # ignoring the argo cd tracking annotation to avoid shared resources with self-heal
    - '.metadata.annotations?."argocd.argoproj.io/tracking-id"'

[nebojsa-prodana]

          resource.customizations.ignoreResourceUpdates.all: |
            jsonPointers:
              - /status
              - /metadata/generation
              - /metadata/resourceVersion
              # ignoring managedFields to prevent churn when argocd competes with argo-rollouts for ownership of vs and dr
              - /metadata/managedFields

          # pods are not directly managed by argocd - skipping churn on larger service (200 or more pods) helps significantly
          resource.customizations.ignoreResourceUpdates.v1_Pod: |
            jsonPointers:
              - /metadata
              - /spec

          # HPA is in control of this, no need to reconcile
          resource.customizations.ignoreResourceUpdates.argoproj.io_Rollout: |
            jsonPointers:
              - /spec/replicas

          # controlled by rollout, no need to reconcile
          resource.customizations.ignoreResourceUpdates.apps_ReplicaSet: |
            jsonPointers:
              - /metadata
              - /spec/replicas

          # controlled by discovery API, no need to reconcile - helps reduce churn with larger services
          resource.customizations.ignoreResourceUpdates.discovery.k8s.io_EndpointSlice:
            |
            jsonPointers:
              - /metadata
              - /endpoints
              - /ports

[agaudreault]

• [Feature request] Ignore resources #9819 (comment) (ReplicaSet annotations)
• Documented configs in operator-manual https://github.com/argoproj/argo-cd/blob/master/docs/operator-manual/argocd-cm.yaml#L114-L134
• Some usage of leases

  resource.customizations.ignoreResourceUpdates._ConfigMap: |
    jqPathExpressions:
    - '.metadata.annotations."control-plane.alpha.kubernetes.io/leader"'
    - .metadata.resourceVersion