ArgoCD is forever stuck on updating the Cluster resource generated by the cluster chart

[spantaleev]

I'm using the following ArgoCD application:

apiVersion: argoproj.io/v1alpha1
kind: Application
metadata:
  name: cloudnative-pg-cluster-something-postgres
  namespace: argocd
  finalizers:
  - resources-finalizer.argocd.argoproj.io
spec:
  destination:
    namespace: something
    name: in-cluster
  source:
    repoURL: 'https://cloudnative-pg.io/charts/'
    targetRevision: 0.1.0
    helm:
      values: |
        fullnameOverride: cluster-postgres

        version:
          postgresql: "17"

        cluster:
          instances: 3

          storage:
            size: 3Gi

          initdb:
            database: something
            owner: something

          enableSuperuserAccess: true
          superuserSecret: cluster-postgres-superuser

          walStorage:
            enabled: true
    chart: cluster
  syncPolicy:
    automated:
      prune: true
      selfHeal: true
  sources: []
  project: common

This generates a Cluster resource with pg_hba: [] and pg_ident: []:

apiVersion: postgresql.cnpg.io/v1
kind: Cluster
metadata:
  labels:
    app.kubernetes.io/instance: cloudnative-pg-cluster-something-postgres
    app.kubernetes.io/managed-by: Helm
    app.kubernetes.io/name: cluster
    app.kubernetes.io/part-of: cloudnative-pg
    argocd.argoproj.io/instance: cloudnative-pg-cluster-something-postgres
    helm.sh/chart: cluster-0.1.0
  name: cluster-postgres
  namespace: something
spec:
  affinity:
    topologyKey: topology.kubernetes.io/zone
  bootstrap:
    initdb:
      database: something
      owner: something
      postInitApplicationSQL: null
  enableSuperuserAccess: true
  imageName: ghcr.io/cloudnative-pg/postgresql:17
  imagePullPolicy: IfNotPresent
  instances: 3
  logLevel: info
  managed: null
  monitoring:
    disableDefaultQueries: false
    enablePodMonitor: true
  postgresGID: 26
  postgresUID: 26
  postgresql:
    parameters: {}
    pg_hba: []
    pg_ident: []
    shared_preload_libraries: null
  primaryUpdateMethod: switchover
  primaryUpdateStrategy: unsupervised
  priorityClassName: null
  storage:
    size: 3Gi
    storageClass: null
  superuserSecret:
    name: cluster-postgres-superuser
  walStorage:
    size: 1Gi
    storageClass: null

ArgoCD continuously retries updating this Cluster resource, but the final result lacks the pg_hba and pg_ident keys. I suspect it's because they are empty lists.

Perhaps the code here could benefit from some if statements that exclude pg_hba and pg_ident if they would end up being empty?

charts/charts/cluster/templates/cluster.yaml

Lines 63 to 66 in 3c5dd5d

	pg_hba:
	{{- toYaml .pg_hba | nindent 6 }}
	pg_ident:
	{{- toYaml .pg_ident | nindent 6 }}

This only became an issue with version 0.1.0 of the cluster chart. Downgrading to 0.0.11 (which does not have these pg_hba and pg_ident keys) fixes the problem.

Seems like pg_hba got introduced in #321 while pg_ident got introduced in #377

[itay-grudev]

We can wrap them in a with statement so they are not included.

[pschichtel]

I'm also seeing this with a bunch of other entries:

I'm using server-side apply

[logan-broit]

We also are running into this issue when using v0.1.0 with fleet, as a workaround we put in values so pg_hba and pg_ident are not empty arrays and that seems to have resolved it for now...

Added this to our values.yaml for our cnpg clusters

    postgresql:
      pg_hba:
        - host all all 10.0.0.0/8 trust
      pg_ident:
        - postgres   local   postgres

Thanks to @cterence for a better idea, adding below to our fleet deployment for the application also temporarily solves the issue

diff:
  comparePatches:
    - apiVersion: postgresql.cnpg.io/v1
      kind: Cluster
      jsonPointers:
        - /spec/postgresql/pg_hba
        - /spec/postgresql/pg_ident

[cterence]

Adding this ignoreDifferences block to the Application temporarily solves the issue

- group: postgresql.cnpg.io
  kind: Cluster
  jqPathExpressions:
    - .spec.postgresql.pg_hba
    - .spec.postgresql.pg_ident

[bozorgmehr96]

I submitted a PR for this, which fixes spec.postgresql.pg_hba, spec.postgresql. pg_ident and spec.postgresql.parameters
If you encounter other keys like this, please let me know so I can add a fix to the PR.