Platform Health

Test file: rossoctl/tests/e2e/openshell/test_01_platform_health.py Tests: 7 | Pass: 7 | Skip: 0 (Kind, fresh cluster)

What This Tests

Validates that core OpenShell platform components (gateway, operator, agent pods) are healthy and running. This is the foundation test that verifies the cluster is in a deployable state before running capability tests.

Architecture Under Test

Test Matrix

Test	Gateway	Operator	weather_agent	adk_agent	claude_sdk_agent	weather_supervised
Pods exist	✅	✅	✅	✅	✅	✅
Pods Running	✅	✅	✅	✅	✅	✅
Containers ready	✅	—	✅	✅	✅	✅
Deployments ready	—	—	✅	✅	✅	✅
No CrashLoopBackOff	—	—	✅	✅	✅	✅

Test Details

test_gateway_pod_running

• What: At least one openshell-gateway pod must be Running
• Asserts: Pod phase == "Running"
• Debug points: Pod name, phase
• Agent coverage: Gateway (openshell-system namespace)

test_gateway_containers_ready

• What: All containers in gateway pod must have ready=true
• Asserts: containerStatuses[*].ready == true
• Debug points: Container names, ready status
• Agent coverage: Gateway (openshell-system namespace)

test_operator_pod_running

• What: At least one rossoctl-operator pod must be Running
• Asserts: Pod phase == "Running"
• Debug points: Pod name, phase
• Agent coverage: Operator (rossoctl-system namespace)
• Skip condition: No operator pods found (operator may not be deployed)

test_all_agent_pods_exist

• What: Each expected agent must have at least one pod
• Asserts: Pod exists for each discovered agent deployment
• Debug points: Agent deployment names, pod names
• Agent coverage: ALL agents (dynamically discovered via label selector)

test_all_agent_pods_running

• What: Every agent pod must be in Running phase
• Asserts: Pod phase == "Running" for all agent pods
• Debug points: Pod name, phase
• Agent coverage: ALL agents (dynamically discovered)

test_agent_deployments_ready

• What: Every agent deployment must have all replicas ready
• Asserts: readyReplicas >= spec.replicas
• Debug points: Deployment name, ready/desired replica counts
• Agent coverage: ALL agents (dynamically discovered)

test_no_crashlooping_agent_pods

• What: No agent pod container should be in CrashLoopBackOff
• Asserts: containerStatuses[*].state.waiting.reason != "CrashLoopBackOff"
• Debug points: Container name, restart count, waiting reason
• Agent coverage: ALL agents (dynamically discovered)

Discovery Mechanism

Tests use dynamic agent discovery instead of hardcoded names:

def _discover_agents(namespace: str) -> list[str]:
    """Discover agent Deployments with readyReplicas > 0."""
    result = kubectl get deploy -n {ns} -l rossoctl.io/type=agent -o json
    return [d["metadata"]["name"] for d in items if d.status.readyReplicas > 0]

This allows the same tests to work on:

• Kind — all 4 agents (custom images available locally)
• HyperShift — only weather-agent (other agents skip if images not in registry)

Future Expansion

No expansion needed — this test is complete and covers all deployment scenarios.

Common Failure Modes

Symptom	Cause	Fix
Gateway pod not Running	Image pull failure	Check gateway image availability
Operator pod not found	Operator not deployed	Deploy rossoctl operator
Agent pod CrashLoopBackOff	Missing LLM secret	Verify LiteLLM virtual keys
Agent deployment not ready	Image pull pending	Wait for image pull or push to registry