Merge pull request #30 from nerdalert/minikube-storage

Replaced broken glusterfs for single node hostPath

Author: tumido

helpers/k8s/model-storage-rwx-pvc-minikube.yaml

@@ -216,17 +216,16 @@ completions are working.
 
 ```bash
 NAMESPACE=llm-d
-POD_IP=$(kubectl get pods -n ${NAMESPACE} -o jsonpath='{range .items[*]}{.metadata.name}{" "}{.status.podIP}{"\n"}{end}' | grep llama-32-3b-instruct-model-service-decode | awk '{print $2}')
-kubectl run --rm -i curl-temp --image=curlimages/curl --restart=Never -- \
-  curl -X POST \
-  "http://${POD_IP}:8000/v1/chat/completions" \
-  -H 'accept: application/json' \
-  -H 'Content-Type: application/json' \
-  -d '{
-    "model": "Llama-3.2-3B-Instruct",
-    "messages": [{"content": "Who won the World Series in 1986?", "role": "user"}],
-    "stream": false
-  }'
+POD=$(kubectl get pods -n "$NAMESPACE" -l llm-d.ai/role=decode -o jsonpath='{.items[0].metadata.name}')
+kubectl exec -n "$NAMESPACE" -c vllm "$POD" -- \
+  curl -sS -X POST http://localhost:8000/v1/chat/completions \
+    -H 'accept: application/json' \
+    -H 'Content-Type: application/json' \
+    -d '{
+      "model": "Llama-32-3B-Instruct",
+      "messages": [{"role":"user","content":"Who won the World Series in 1986?"}],
+      "stream": false
+    }'
 ```
 
 After the p/d pods are running, you can view the models being run on the GPUs on the host to verify activity.
@@ -248,3 +247,9 @@ make a change, simply uninstall and then run the installer again with any change
 ```bash
 ./llmd-installer.sh --uninstall
 ```
+
+To remove the minikube cluster this simply wraps the minikube command for convenience.
+
+```bash
+./llmd-installer.sh --delete-minikube
+```

quickstart/README-minikube.md

@@ -8,7 +8,7 @@ NAMESPACE="llm-d"
 PROVISION_MINIKUBE=false
 PROVISION_MINIKUBE_GPU=false
 USE_MINIKUBE_STORAGE=false
-STORAGE_SIZE="7Gi"
+STORAGE_SIZE="15Gi"
 STORAGE_CLASS="efs-sc"
 DELETE_MINIKUBE=false
 ACTION="install"
@@ -158,7 +158,7 @@ validate_hf_token() {
 ### MINIKUBE HANDLERS ###
 provision_minikube() {
   log_info "🌱 Provisioning Minikube cluster..."
-  minikube start --nodes=3
+  minikube start
   log_success "🚀 Minikube started."
 }
 
@@ -167,8 +167,7 @@ provision_minikube_gpu() {
   minikube start \
     --driver docker \
     --container-runtime docker \
-    --gpus all \
-    --nodes=3
+    --gpus all
   log_success "🚀 Minikube GPU cluster started."
 }
 
@@ -248,10 +247,8 @@ install() {
   fi
   log_success "✅ Job manifest patched"
 
-  # This is where all minicube scenarios use glusterfs sc and existing kube clusters use model-storage-rwx-pvc.yaml
   log_info "💾 Provisioning model storage…"
   if [[ "${USE_MINIKUBE_STORAGE}" == "true" ]]; then
-    ensure_gluster_addon_enabled
     # this creates both the hostPath PV and the matching PVC
     setup_minikube_storage
     log_success "✅ PVC created from model-storage-rwx-pvc-minikube.yaml"
@@ -307,7 +304,7 @@ kind: PersistentVolume
 metadata:
   name: redis-hostpath-pv
 spec:
-  storageClassName: glusterfile
+  storageClassName: manual
   capacity:
     storage: 5Gi
   accessModes:
@@ -320,10 +317,10 @@ spec:
 apiVersion: v1
 kind: PersistentVolumeClaim
 metadata:
-  name: redis-pvc
+  name: redis-data-redis-master
   namespace: ${NAMESPACE}
 spec:
-  storageClassName: glusterfile
+  storageClassName: manual
   accessModes:
     - ReadWriteMany
   resources:
@@ -340,17 +337,17 @@ EOF
 }
 
 setup_minikube_storage() {
-  log_info "📦 Setting up Minikube Gluster RWX Shared Storage..."
+  log_info "📦 Setting up Minikube hostPath RWX Shared Storage..."
   log_info "🔄 Creating PV and PVC for llama model..."
 kubectl apply -f - <<EOF
 apiVersion: v1
 kind: PersistentVolume
 metadata:
   name: llama-hostpath-pv
 spec:
-  storageClassName: glusterfile
+  storageClassName: manual
   capacity:
-    storage: 7Gi
+    storage: ${STORAGE_SIZE}
   accessModes:
     - ReadWriteMany
   persistentVolumeReclaimPolicy: Retain
@@ -364,30 +361,17 @@ metadata:
   name: llama-3.2-3b-instruct-pvc
   namespace: ${NAMESPACE}
 spec:
-  storageClassName: glusterfile
+  storageClassName: manual
   accessModes:
     - ReadWriteMany
   resources:
     requests:
-      storage: 7Gi
+      storage: ${STORAGE_SIZE}
   volumeName: llama-hostpath-pv
 EOF
   log_success "✅ llama model PV and PVC created."
 }
 
-ensure_gluster_addon_enabled() {
-  if [[ "${USE_MINIKUBE_STORAGE}" == "true" ]]; then
-    log_info "🔍 Checking if Gluster addon is enabled in Minikube..."
-    if ! minikube addons list | grep -q 'storage-provisioner-gluster.*disabled'; then
-      log_success "✅ Gluster addon already enabled."
-    else
-      log_info "📦 Enabling Gluster addon for Minikube..."
-      minikube addons enable storage-provisioner-gluster
-      log_success "✅ Gluster addon enabled."
-    fi
-  fi
-}
-
 clone_gaie_repo() {
   if [[ ! -d gateway-api-inference-extension ]]; then
     git clone https://github.com/neuralmagic/gateway-api-inference-extension.git