Guides on contineroo.ch

Monitoring with Prometheus

Mon, 01 Jan 0001 00:00:00 +0000

Prerequisites

Prometheus Operator or kube-prometheus-stack.
agent-forge-operator installed through the Helm chart.

Enable metrics resources

Create a values.yaml file:

---
metrics:
  serviceMonitor:
    enabled: true
  prometheusRule:
    enabled: true

Install or upgrade the Helm chart by following the installation guide.

Available metrics

The controller exposes these application metrics:

Metric	Meaning
`agent_forge_vsphere_vm_operations_total`	VM create/delete attempts by operation and result.
`agent_forge_iso_operations_total`	ISO cache ensure/delete attempts by operation and result.
`agent_forge_pool_capacity`	Per-pool capacity gauges for desired, waiting, available, pending, and planned-create counts.

You can also use kube-state-metrics to alert on VsphereAgentPool and VsphereAgent conditions once those custom resources are configured for state metrics collection.

Troubleshooting

Mon, 01 Jan 0001 00:00:00 +0000

AgentMachine demand is not observed

Check AgentMachines in the hosted control plane namespace:

kubectl -n demo-demo get agentmachines.capi-provider.agent-install.openshift.io -o yaml
kubectl -n demo get vsphereagentpool demo-worker \
  -o jsonpath='{.status.conditions[?(@.type=="AgentMachineDemandFound")]}{"\n"}'

Confirm that the AgentMachine has the expected hypershift.openshift.io/nodePool annotation and reports Ready=False with Reason=NoSuitableAgents.

InfraEnv ISO is unavailable

kubectl -n demo get infraenv demo -o yaml
kubectl -n demo get vsphereagentpool demo-worker \
  -o jsonpath='{.status.conditions[?(@.type=="InfraEnvAvailable")]}{"\n"}'

The operator cannot create bootable discovery VMs until status.isoDownloadURL is populated on the InfraEnv.

Cleanup Policy

Mon, 01 Jan 0001 00:00:00 +0000

spec.cleanupPolicy controls whether agent-forge-operator deletes external inventory when demand disappears or when a pool is deleted.

Delete

Delete is the default policy.

spec:
  cleanupPolicy: Delete

With Delete, the operator removes stale owned vSphere VMs and unbound Agents when they are no longer needed. Scale-down is conservative: the controller first observes a paired CAPI Machine entering deletion, waits until that Machine has disappeared, and only then deletes the paired VsphereAgent, VM, and stale Agent.