Metricbeat: Dashboard desired pods shown as 0 (zero)

[mgo1 (Mauricio)]

Hi,
As the title suggests I'm getting some metrics in the [Metricbeat Kubernetes] ECS Overview but I have a couple of issues.

1. The desired pods and available pods is always zero (even though there are around 70 pods running in my cluster).
2. Metrics are shown intermittently: sometimes I get deployments, nodes, memory usage by pod etc. But sometimes everything is zero. I have to play with the time range, and autorefresh to get something.

My configuration
Kubernetes Cluster version 1.21.2 running on Azure (AKS)
ECK version 1.7
Elasticsearch, kibana and metricbeats version 7.14.0
I'm using kube-metrics v2.2.0 in kube-system namespace
This is the configuration i'm using
https://raw.githubusercontent.com/elastic/cloud-on-k8s/1.7/config/recipes/beats/metricbeat_hosts.yaml

I added the following module

      - module: kubernetes
        period: 10s
        hosts:
        - kube-state-metrics.kube-system.svc.cluster.local:8080
        metricsets:
        - state_node
        - state_deployment
        - state_replicaset
        - state_statefulset
        - state_pod
        - state_container

The logs show no errors, and all pods related from Elastic stack are running in their own namespace.

Any idea what could be causing this? Thanks for your help.