docs(demos): expand cloud burst resilience example

demos/grid-cloud-burst/example-manifests/08-regional-local-pools.yaml

@@ -0,0 +1,202 @@
+# Two local llm-d-inference-sim providers per regional gateway.
+# Names are intentionally shared across Deployment, Service, InferenceProvider,
+# and provider-gateway load_balancer clusters.
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: llm-d-east-1-config
+  namespace: grid-system
+data:
+  config.yaml: |
+    port: 8000
+    model: "gpt-4o-mini"
+    served-model-name: ["gpt-4o-mini"]
+    mode: "echo"
+    force-dummy-tokenizer: true
+    max-num-seqs: 4
+    max-waiting-queue-length: 128
+    fake-metrics:
+      running-requests: 0
+      waiting-requests: 0
+      kv-cache-usage: 0.1
+---
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: llm-d-east-2-config
+  namespace: grid-system
+data:
+  config.yaml: |
+    port: 8000
+    model: "gpt-4o-mini"
+    served-model-name: ["gpt-4o-mini"]
+    mode: "echo"
+    force-dummy-tokenizer: true
+    max-num-seqs: 4
+    max-waiting-queue-length: 128
+    fake-metrics:
+      running-requests: 0
+      waiting-requests: 0
+      kv-cache-usage: 0.1
+---
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: llm-d-west-1-config
+  namespace: grid-system
+data:
+  config.yaml: |
+    port: 8000
+    model: "gpt-4o-mini"
+    served-model-name: ["gpt-4o-mini"]
+    mode: "echo"
+    force-dummy-tokenizer: true
+    max-num-seqs: 4
+    max-waiting-queue-length: 128
+    fake-metrics:
+      running-requests: 0
+      waiting-requests: 0
+      kv-cache-usage: 0.1
+---
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: llm-d-west-2-config
+  namespace: grid-system
+data:
+  config.yaml: |
+    port: 8000
+    model: "gpt-4o-mini"
+    served-model-name: ["gpt-4o-mini"]
+    mode: "echo"
+    force-dummy-tokenizer: true
+    max-num-seqs: 4
+    max-waiting-queue-length: 128
+    fake-metrics:
+      running-requests: 0
+      waiting-requests: 0
+      kv-cache-usage: 0.1
+---
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: llm-d-east-1
+  namespace: grid-system
+  labels: {app.kubernetes.io/name: llm-d-inference-sim, grid.praxis-proxy.io/pool: llm-d-east-1}
+spec:
+  replicas: 1
+  selector: {matchLabels: {grid.praxis-proxy.io/pool: llm-d-east-1}}
+  template:
+    metadata: {labels: {app.kubernetes.io/name: llm-d-inference-sim, grid.praxis-proxy.io/pool: llm-d-east-1}}
+    spec:
+      automountServiceAccountToken: false
+      securityContext: {runAsNonRoot: true, seccompProfile: {type: RuntimeDefault}}
+      containers:
+      - name: simulator
+        image: ghcr.io/llm-d/llm-d-inference-sim:v0.10.2
+        imagePullPolicy: IfNotPresent
+        args: ["--config", "/config/config.yaml"]
+        ports: [{name: http, containerPort: 8000}]
+        volumeMounts: [{name: config, mountPath: /config, readOnly: true}]
+        readinessProbe: {httpGet: {path: /health, port: http}, periodSeconds: 5}
+        securityContext: {allowPrivilegeEscalation: false, capabilities: {drop: [ALL]}, readOnlyRootFilesystem: true}
+        resources: {requests: {cpu: 100m, memory: 128Mi}, limits: {cpu: "1", memory: 512Mi}}
+      volumes: [{name: config, configMap: {name: llm-d-east-1-config}}]
+---
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: llm-d-east-2
+  namespace: grid-system
+  labels: {app.kubernetes.io/name: llm-d-inference-sim, grid.praxis-proxy.io/pool: llm-d-east-2}
+spec:
+  replicas: 1
+  selector: {matchLabels: {grid.praxis-proxy.io/pool: llm-d-east-2}}
+  template:
+    metadata: {labels: {app.kubernetes.io/name: llm-d-inference-sim, grid.praxis-proxy.io/pool: llm-d-east-2}}
+    spec:
+      automountServiceAccountToken: false
+      securityContext: {runAsNonRoot: true, seccompProfile: {type: RuntimeDefault}}
+      containers:
+      - name: simulator
+        image: ghcr.io/llm-d/llm-d-inference-sim:v0.10.2
+        imagePullPolicy: IfNotPresent
+        args: ["--config", "/config/config.yaml"]
+        ports: [{name: http, containerPort: 8000}]
+        volumeMounts: [{name: config, mountPath: /config, readOnly: true}]
+        readinessProbe: {httpGet: {path: /health, port: http}, periodSeconds: 5}
+        securityContext: {allowPrivilegeEscalation: false, capabilities: {drop: [ALL]}, readOnlyRootFilesystem: true}
+        resources: {requests: {cpu: 100m, memory: 128Mi}, limits: {cpu: "1", memory: 512Mi}}
+      volumes: [{name: config, configMap: {name: llm-d-east-2-config}}]
+---
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: llm-d-west-1
+  namespace: grid-system
+  labels: {app.kubernetes.io/name: llm-d-inference-sim, grid.praxis-proxy.io/pool: llm-d-west-1}
+spec:
+  replicas: 1
+  selector: {matchLabels: {grid.praxis-proxy.io/pool: llm-d-west-1}}
+  template:
+    metadata: {labels: {app.kubernetes.io/name: llm-d-inference-sim, grid.praxis-proxy.io/pool: llm-d-west-1}}
+    spec:
+      automountServiceAccountToken: false
+      securityContext: {runAsNonRoot: true, seccompProfile: {type: RuntimeDefault}}
+      containers:
+      - name: simulator
+        image: ghcr.io/llm-d/llm-d-inference-sim:v0.10.2
+        imagePullPolicy: IfNotPresent
+        args: ["--config", "/config/config.yaml"]
+        ports: [{name: http, containerPort: 8000}]
+        volumeMounts: [{name: config, mountPath: /config, readOnly: true}]
+        readinessProbe: {httpGet: {path: /health, port: http}, periodSeconds: 5}
+        securityContext: {allowPrivilegeEscalation: false, capabilities: {drop: [ALL]}, readOnlyRootFilesystem: true}
+        resources: {requests: {cpu: 100m, memory: 128Mi}, limits: {cpu: "1", memory: 512Mi}}
+      volumes: [{name: config, configMap: {name: llm-d-west-1-config}}]
+---
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: llm-d-west-2
+  namespace: grid-system
+  labels: {app.kubernetes.io/name: llm-d-inference-sim, grid.praxis-proxy.io/pool: llm-d-west-2}
+spec:
+  replicas: 1
+  selector: {matchLabels: {grid.praxis-proxy.io/pool: llm-d-west-2}}
+  template:
+    metadata: {labels: {app.kubernetes.io/name: llm-d-inference-sim, grid.praxis-proxy.io/pool: llm-d-west-2}}
+    spec:
+      automountServiceAccountToken: false
+      securityContext: {runAsNonRoot: true, seccompProfile: {type: RuntimeDefault}}
+      containers:
+      - name: simulator
+        image: ghcr.io/llm-d/llm-d-inference-sim:v0.10.2
+        imagePullPolicy: IfNotPresent
+        args: ["--config", "/config/config.yaml"]
+        ports: [{name: http, containerPort: 8000}]
+        volumeMounts: [{name: config, mountPath: /config, readOnly: true}]
+        readinessProbe: {httpGet: {path: /health, port: http}, periodSeconds: 5}
+        securityContext: {allowPrivilegeEscalation: false, capabilities: {drop: [ALL]}, readOnlyRootFilesystem: true}
+        resources: {requests: {cpu: 100m, memory: 128Mi}, limits: {cpu: "1", memory: 512Mi}}
+      volumes: [{name: config, configMap: {name: llm-d-west-2-config}}]
+---
+apiVersion: v1
+kind: Service
+metadata: {name: llm-d-east-1, namespace: grid-system}
+spec: {selector: {grid.praxis-proxy.io/pool: llm-d-east-1}, ports: [{name: http, port: 8000, targetPort: http}]}
+---
+apiVersion: v1
+kind: Service
+metadata: {name: llm-d-east-2, namespace: grid-system}
+spec: {selector: {grid.praxis-proxy.io/pool: llm-d-east-2}, ports: [{name: http, port: 8000, targetPort: http}]}
+---
+apiVersion: v1
+kind: Service
+metadata: {name: llm-d-west-1, namespace: grid-system}
+spec: {selector: {grid.praxis-proxy.io/pool: llm-d-west-1}, ports: [{name: http, port: 8000, targetPort: http}]}
+---
+apiVersion: v1
+kind: Service
+metadata: {name: llm-d-west-2, namespace: grid-system}
+spec: {selector: {grid.praxis-proxy.io/pool: llm-d-west-2}, ports: [{name: http, port: 8000, targetPort: http}]}

demos/grid-cloud-burst/example-manifests/09-regional-inference-providers.yaml

@@ -0,0 +1,65 @@
+# Two local InferenceProviders per site. Use localityScope: sameSite on the
+# GridNetwork so east and west remain isolated active groups.
+apiVersion: grid.praxis-proxy.io/v1alpha1
+kind: InferenceProvider
+metadata: {name: llm-d-east-1, namespace: grid-system}
+spec:
+  gridNetworkRef: grid-cloud-burst-rhoai
+  providerKind: llm-d-inference-sim
+  backendKind: local
+  endpoint: http://llm-d-east-1.grid-system.svc.cluster.local:8000
+  routingClusterRef: llm-d-east-1
+  capacityWeight: 50
+  models: [{name: gpt-4o-mini, capabilities: [text_generation], contextWindow: 4096}]
+  siteSelector: {matchLabels: {grid.praxis-proxy.io/provider-site: east-local}}
+  accessPolicy: {siteSelector: {matchLabels: {}}}
+  metricsConfig: {metricsEndpoint: http://llm-d-east-1.grid-system.svc.cluster.local:8000, path: /metrics, queueCapacity: 10, signalNames: {queueDepth: vllm:num_requests_waiting}, timeout: 2s}
+  healthCheck: {path: /health, interval: 5s, timeout: 2s}
+---
+apiVersion: grid.praxis-proxy.io/v1alpha1
+kind: InferenceProvider
+metadata: {name: llm-d-east-2, namespace: grid-system}
+spec:
+  gridNetworkRef: grid-cloud-burst-rhoai
+  providerKind: llm-d-inference-sim
+  backendKind: local
+  endpoint: http://llm-d-east-2.grid-system.svc.cluster.local:8000
+  routingClusterRef: llm-d-east-2
+  capacityWeight: 50
+  models: [{name: gpt-4o-mini, capabilities: [text_generation], contextWindow: 4096}]
+  siteSelector: {matchLabels: {grid.praxis-proxy.io/provider-site: east-local}}
+  accessPolicy: {siteSelector: {matchLabels: {}}}
+  metricsConfig: {metricsEndpoint: http://llm-d-east-2.grid-system.svc.cluster.local:8000, path: /metrics, queueCapacity: 10, signalNames: {queueDepth: vllm:num_requests_waiting}, timeout: 2s}
+  healthCheck: {path: /health, interval: 5s, timeout: 2s}
+---
+apiVersion: grid.praxis-proxy.io/v1alpha1
+kind: InferenceProvider
+metadata: {name: llm-d-west-1, namespace: grid-system}
+spec:
+  gridNetworkRef: grid-cloud-burst-rhoai
+  providerKind: llm-d-inference-sim
+  backendKind: local
+  endpoint: http://llm-d-west-1.grid-system.svc.cluster.local:8000
+  routingClusterRef: llm-d-west-1
+  capacityWeight: 50
+  models: [{name: gpt-4o-mini, capabilities: [text_generation], contextWindow: 4096}]
+  siteSelector: {matchLabels: {grid.praxis-proxy.io/provider-site: west-local}}
+  accessPolicy: {siteSelector: {matchLabels: {}}}
+  metricsConfig: {metricsEndpoint: http://llm-d-west-1.grid-system.svc.cluster.local:8000, path: /metrics, queueCapacity: 10, signalNames: {queueDepth: vllm:num_requests_waiting}, timeout: 2s}
+  healthCheck: {path: /health, interval: 5s, timeout: 2s}
+---
+apiVersion: grid.praxis-proxy.io/v1alpha1
+kind: InferenceProvider
+metadata: {name: llm-d-west-2, namespace: grid-system}
+spec:
+  gridNetworkRef: grid-cloud-burst-rhoai
+  providerKind: llm-d-inference-sim
+  backendKind: local
+  endpoint: http://llm-d-west-2.grid-system.svc.cluster.local:8000
+  routingClusterRef: llm-d-west-2
+  capacityWeight: 50
+  models: [{name: gpt-4o-mini, capabilities: [text_generation], contextWindow: 4096}]
+  siteSelector: {matchLabels: {grid.praxis-proxy.io/provider-site: west-local}}
+  accessPolicy: {siteSelector: {matchLabels: {}}}
+  metricsConfig: {metricsEndpoint: http://llm-d-west-2.grid-system.svc.cluster.local:8000, path: /metrics, queueCapacity: 10, signalNames: {queueDepth: vllm:num_requests_waiting}, timeout: 2s}
+  healthCheck: {path: /health, interval: 5s, timeout: 2s}

demos/grid-cloud-burst/example-manifests/10-observability-jaeger.yaml

@@ -0,0 +1,67 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: jaeger
+  namespace: grid-system
+  labels:
+    app.kubernetes.io/name: jaeger
+    app.kubernetes.io/part-of: grid-cloud-burst
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app.kubernetes.io/name: jaeger
+  template:
+    metadata:
+      labels:
+        app.kubernetes.io/name: jaeger
+        app.kubernetes.io/part-of: grid-cloud-burst
+    spec:
+      containers:
+        - name: jaeger
+          image: jaegertracing/all-in-one:1.57
+          imagePullPolicy: IfNotPresent
+          env:
+            - name: COLLECTOR_OTLP_ENABLED
+              value: "true"
+          ports:
+            - name: otlp-grpc
+              containerPort: 4317
+            - name: query
+              containerPort: 16686
+          readinessProbe:
+            httpGet:
+              path: /
+              port: query
+            initialDelaySeconds: 5
+            periodSeconds: 5
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: jaeger-collector
+  namespace: grid-system
+  labels:
+    app.kubernetes.io/name: jaeger
+spec:
+  selector:
+    app.kubernetes.io/name: jaeger
+  ports:
+    - name: otlp-grpc
+      port: 4317
+      targetPort: otlp-grpc
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: jaeger-query
+  namespace: grid-system
+  labels:
+    app.kubernetes.io/name: jaeger
+spec:
+  selector:
+    app.kubernetes.io/name: jaeger
+  ports:
+    - name: query
+      port: 16686
+      targetPort: query

demos/grid-cloud-burst/example-manifests/10-openai-overflow-providers.yaml

@@ -0,0 +1,53 @@
+# OpenAI is an overflow provider. Its InferenceProvider endpoint must describe
+# the real external backend; the provider-gateway route separately owns the
+# final-hop TLS/authority and credential injection configuration. The paired
+# provider-gateway ConfigMaps must set both tls.sni and http.authority to
+# api.openai.com; SNI alone can produce HTTP 421 from the upstream. In Praxis
+# configuration authority is nested under the cluster's http field.
+apiVersion: grid.praxis-proxy.io/v1alpha1
+kind: InferenceProvider
+metadata:
+  name: openai-east
+  namespace: grid-system
+spec:
+  gridNetworkRef: grid-cloud-burst-rhoai
+  backendKind: api_provider
+  providerKind: openai
+  endpoint: https://api.openai.com
+  routingClusterRef: openai-east
+  capacityWeight: 50
+  models:
+    - name: gpt-4o-mini
+      capabilities: [text_generation]
+      contextWindow: 4096
+  auth:
+    strategy: bearer_token
+    manual: true
+    secretRef: {name: openai-api-key, namespace: grid-system, key: token}
+  siteSelector:
+    matchLabels: {grid.praxis-proxy.io/provider-site: east-cloud}
+  accessPolicy: {siteSelector: {matchLabels: {}}}
+---
+apiVersion: grid.praxis-proxy.io/v1alpha1
+kind: InferenceProvider
+metadata:
+  name: openai-west
+  namespace: grid-system
+spec:
+  gridNetworkRef: grid-cloud-burst-rhoai
+  backendKind: api_provider
+  providerKind: openai
+  endpoint: https://api.openai.com
+  routingClusterRef: openai-west
+  capacityWeight: 50
+  models:
+    - name: gpt-4o-mini
+      capabilities: [text_generation]
+      contextWindow: 4096
+  auth:
+    strategy: bearer_token
+    manual: true
+    secretRef: {name: openai-api-key, namespace: grid-system, key: token}
+  siteSelector:
+    matchLabels: {grid.praxis-proxy.io/provider-site: west-cloud}
+  accessPolicy: {siteSelector: {matchLabels: {}}}