multi tenant patches and guards (#1049)

* multi tenant patches and guards

Signed-off-by: Mohammed Abdi <mohammed.munir.abdi@ibm.com>

* simplify by removing guardrails

Signed-off-by: Mohammed Abdi <mohammed.munir.abdi@ibm.com>

* update

Signed-off-by: Mohammed Abdi <mohammed.munir.abdi@ibm.com>

* simplify e2e

Signed-off-by: Mohammed Abdi <mohammed.munir.abdi@ibm.com>

* address review

Signed-off-by: Mohammed Abdi <mohammed.munir.abdi@ibm.com>

* fix ci

Signed-off-by: Mohammed Abdi <mohammed.munir.abdi@ibm.com>

---------

Signed-off-by: Mohammed Abdi <mohammed.munir.abdi@ibm.com>

Author: mamy-CS

.github/workflows/ci-pr-checks.yaml

@@ -324,6 +324,8 @@ jobs:
           CREATE_CLUSTER: "true"
           INSTALL_GATEWAY_CTRLPLANE: "true"
           E2E_TESTS_ENABLED: "true"
+          # Dual-controller smoke installs a second Helm release; chart path must be explicit in CI.
+          WVA_E2E_CHART_PATH: ${{ github.workspace }}/charts/workload-variant-autoscaler
           IMG: ${{ steps.build-image.outputs.image }}
           SKIP_BUILD: "true"
           PROMETHEUS_ADAPTER_WAIT: "false"

Makefile

@@ -25,6 +25,7 @@ SCALE_TO_ZERO_ENABLED       ?= false
 SCALER_BACKEND              ?= prometheus-adapter  # prometheus-adapter (HPA), keda (ScaledObject), or none (skip, use pre-installed backend)
 E2E_MONITORING_NAMESPACE    ?= workload-variant-autoscaler-monitoring
 E2E_EMULATED_LLMD_NAMESPACE ?= llm-d-sim
+E2E_WVA_CHART_PATH          ?= $(CURDIR)/charts/workload-variant-autoscaler
 BENCHMARK_SCENARIO          ?= prefill_heavy  # Options: prefill_heavy (phase3a), decode_heavy (decode-heavy)
 
 # Map scenario name to Ginkgo label filter
@@ -328,11 +329,12 @@ test-e2e-smoke: ## Run smoke e2e tests
 	WVA_NAMESPACE=$(CONTROLLER_NAMESPACE) \
 	LLMD_NAMESPACE=$(E2E_EMULATED_LLMD_NAMESPACE) \
 	MONITORING_NAMESPACE=$(E2E_MONITORING_NAMESPACE) \
+	WVA_E2E_CHART_PATH=$${WVA_E2E_CHART_PATH:-$(E2E_WVA_CHART_PATH)} \
 	USE_SIMULATOR=$(USE_SIMULATOR) \
 	SCALE_TO_ZERO_ENABLED=$(SCALE_TO_ZERO_ENABLED) \
 	SCALER_BACKEND=$(SCALER_BACKEND) \
 	MODEL_ID=$(MODEL_ID) \
-	go test ./test/e2e/ -timeout 20m -v -ginkgo.v \
+	go test ./test/e2e/ -timeout 35m -v -ginkgo.v \
 		-ginkgo.label-filter="smoke" $(FOCUS_ARGS) $(SKIP_ARGS); \
 	TEST_EXIT_CODE=$$?; \
 	echo ""; \
@@ -350,6 +352,7 @@ test-e2e-full: ## Run full e2e test suite
 	KUBECONFIG=$(KUBECONFIG) \
 	ENVIRONMENT=$(ENVIRONMENT) \
 	WVA_NAMESPACE=$(CONTROLLER_NAMESPACE) \
+	WVA_E2E_CHART_PATH=$${WVA_E2E_CHART_PATH:-$(E2E_WVA_CHART_PATH)} \
 	USE_SIMULATOR=$(USE_SIMULATOR) \
 	SCALE_TO_ZERO_ENABLED=$(SCALE_TO_ZERO_ENABLED) \
 	SCALER_BACKEND=$(SCALER_BACKEND) \

config/samples/hpa/hpa.yaml

@@ -31,6 +31,7 @@ spec:
         selector:
           matchLabels:
             variant_name: sample-deployment
+            exported_namespace: llm-d-sim
       target:
         type: AverageValue
         averageValue: "1"

deploy/README.md

@@ -586,6 +586,7 @@ spec:
         selector:
           matchLabels:
             variant_name: my-vllm-deployment-decode
+            exported_namespace: $NAMESPACE
       target:
         type: AverageValue
         averageValue: "1"

docs/user-guide/hpa-integration.md

@@ -523,6 +523,7 @@ spec:
         selector:
           matchLabels:
             variant_name: sample-deployment
+            exported_namespace: llm-d-sim
       target:
         type: AverageValue
         averageValue: "1"

docs/user-guide/multi-controller-isolation.md

@@ -219,6 +219,7 @@ spec:
         selector:
           matchLabels:
             variant_name: "llama-8b"
+            exported_namespace: "llm-d"
             controller_instance: "my-instance-id"
 ```
 

test/e2e/fixtures/hpa_builder.go

@@ -114,7 +114,10 @@ func buildHPA(namespace, name, deploymentName, vaName string, minReplicas, maxRe
 						Metric: autoscalingv2.MetricIdentifier{
 							Name: "wva_desired_replicas",
 							Selector: &metav1.LabelSelector{
-								MatchLabels: map[string]string{"variant_name": vaName},
+								MatchLabels: map[string]string{
+									"variant_name":       vaName,
+									"exported_namespace": namespace,
+								},
 							},
 						},
 						Target: autoscalingv2.MetricTarget{

test/e2e/smoke_test.go

@@ -944,7 +944,8 @@ func CreateHPAOnDesiredReplicaMetrics(name, namespace, deploymentName, variantNa
 							Name: constants.WVADesiredReplicas,
 							Selector: &metav1.LabelSelector{
 								MatchLabels: map[string]string{
-									"variant_name": variantName,
+									"variant_name":       variantName,
+									"exported_namespace": namespace,
 								},
 							},
 						},