Problem
The embedded MCP server in OLS currently uses a hardcoded toolset list: ["core", "config", "helm", "metrics"]. The kubevirt toolset is not enabled even when OpenShift Virtualization is installed on the cluster.
This means users with OpenShift Virtualization get no VM-specific troubleshooting tools in OLS unless an admin manually patches the openshift-mcp-server-config ConfigMap — which the operator immediately reverts on reconciliation.
Proposal
The OLS operator should auto-detect whether OpenShift Virtualization is installed (e.g., by checking for the HyperConverged CRD or the kubevirt-hyperconverged CR in openshift-cnv namespace) and automatically add "kubevirt" to the MCP server toolsets during ConfigMap reconciliation.
Detection logic (example)
// During reconcileAppServerResources:
if crdExists("hyperconvergeds.hco.kubevirt.io") {
toolsets = append(toolsets, "kubevirt")
}
Why this matters
We tested OLS with the kubevirt toolset enabled against three VM troubleshooting scenarios (storage failure, crashloop, migration failure). With gpt-5.2 in troubleshooting mode, OLS correctly identified root causes in all three cases using the kubevirt MCP tools. Without the toolset enabled, OLS can only use alert data and generic Kubernetes tools, producing significantly weaker diagnostics.
Testing results
| Model |
kubevirt toolset |
Root causes identified |
| gpt-4o-mini |
disabled |
0/3 (only alert summaries) |
| gpt-4o-mini |
enabled + explicit prompt |
2/3 |
| gpt-5.2 |
enabled |
3/3 (all scenarios, simple questions) |
Related PRs
Acceptance criteria
- When OpenShift Virtualization is installed, the kubevirt toolset is automatically enabled in the MCP server config
- When OpenShift Virtualization is NOT installed, the kubevirt toolset is not included (no error, no degradation)
- No manual configuration required by the cluster admin
Problem
The embedded MCP server in OLS currently uses a hardcoded toolset list:
["core", "config", "helm", "metrics"]. Thekubevirttoolset is not enabled even when OpenShift Virtualization is installed on the cluster.This means users with OpenShift Virtualization get no VM-specific troubleshooting tools in OLS unless an admin manually patches the
openshift-mcp-server-configConfigMap — which the operator immediately reverts on reconciliation.Proposal
The OLS operator should auto-detect whether OpenShift Virtualization is installed (e.g., by checking for the
HyperConvergedCRD or thekubevirt-hyperconvergedCR inopenshift-cnvnamespace) and automatically add"kubevirt"to the MCP server toolsets during ConfigMap reconciliation.Detection logic (example)
Why this matters
We tested OLS with the kubevirt toolset enabled against three VM troubleshooting scenarios (storage failure, crashloop, migration failure). With
gpt-5.2in troubleshooting mode, OLS correctly identified root causes in all three cases using the kubevirt MCP tools. Without the toolset enabled, OLS can only use alert data and generic Kubernetes tools, producing significantly weaker diagnostics.Testing results
Related PRs
vm_troubleshootas an MCP ToolAcceptance criteria