1.5.0 reconciliation / new cluster fails with `selector` does not match template `labels`

[cayla]

Hi,

I upgraded to 1.5.0 yesterday and since then, I am getting reconciliation errors and I cannot create new clusters.

Existing cluster example:

# Source: pulse/templates/dragonflydb.yaml
apiVersion: dragonflydb.io/v1alpha1
kind: Dragonfly
metadata:
  name: pm-switch-out-ai-api-dragonflydb
  namespace: pulse-stg
  labels:
    app.kubernetes.io/name: pulse
    app.kubernetes.io/instance: pm-switch-out-ai-api
    app.kubernetes.io/component: dragonflydb
    app.kubernetes.io/role: cache
spec:
  args:
    - --cache_mode=true
    - --proactor_threads=3
    - --maxmemory=800mb
  image: docker.dragonFlyDB.io/dragonflydb/dragonfly:v1.37.0
  labels:
    app.kubernetes.io/name: pulse
    app.kubernetes.io/instance: pm-switch-out-ai-api
    app.kubernetes.io/component: dragonflydb
    app.kubernetes.io/role: cache
  replicas: 1
  resources:
    limits:
      ephemeral-storage: 2Gi
      memory: 1Gi
    requests:
      cpu: 500m
      ephemeral-storage: 50Mi
      memory: 1Gi

Error:

manager 2026-03-22T20:33:19Z    ERROR    Reconciler error    {"controller": "Dragonfly", "controllerGroup": "dragonflydb.io", "controllerKind": "Dragonfly", "Dragonfly": {"name":"repro-dragonflydb","namespace":"dragonfly-repro"}, "namespace": "dragonfly-repro", "name": "repro-dragonflydb", "reconcileID": "c074cc5f-7ae4-4698-a
eb5-cf6fd58441f7", "error": "failed to reconcile dragonfly resources: failed to create resource: StatefulSet.apps \"repro-dragonflydb\" is invalid: spec.template.metadata.labels: Invalid value: {\"app\":\"repro-dragonflydb\",\"app.kubernetes.io/name\":\"myapp\",\"app.kubernetes.io/part-of\":\"dragonfly\"}: `selector` does not match template `labels`"}

With selector:

{
  "app": "pm-switch-out-ai-api-dragonflydb",
  "app.kubernetes.io/name": "dragonfly",
  "app.kubernetes.io/part-of": "dragonfly"
}

I also noticed new clusters fail to create. A minimal reproduction:

# repro-dragonfly.yaml
apiVersion: dragonflydb.io/v1alpha1
kind: Dragonfly
metadata:
  name: repro-dragonflydb
  namespace: dragonfly-repro
  labels:
    app.kubernetes.io/name: myapp # metadata labels don't matter
spec:
  replicas: 1
  image: docker.dragonflydb.io/dragonflydb/dragonfly:v1.37.0
  labels:
    app.kubernetes.io/name: myapp # this is the trigger — overrides the hardcoded "dragonfly"

If I comment out the last two lines, it creates fine and it has the selectors:

{
  "app": "repro-dragonflydb",
  "app.kubernetes.io/name": "dragonfly",
  "app.kubernetes.io/part-of": "dragonfly"
}

I believe this was caused by #480. I apologize if I missed any required changes needed on my end.

It went away when I went back to 1.4.0.

Thank you for all your hard work. We have been happily trucking otherwise for sometime now. A great improvement over memcache and redis.

PS - and yes, I now see that my the label app.kubernetes.io/name was never applying cleanly from the CR in the first place and always dragonfly, but still thought this was worth reporting.

EDIT: I dropped the .spec.labels.app.kubernetes.io/name in our CRs so we could unblock the upgrade, but I am curious to understand if #480 was an incomplete fix and this is a bug or not.

[cthtrifork]

I am experiencing the same issue

[Abhra303]

Hey, sorry for the bug, I will fix it.