last

gateway.yaml

@@ -0,0 +1,38 @@
+# Gateway.yaml
+apiVersion: networking.istio.io/v1alpha3
+kind: Gateway
+metadata:
+  name: ssh2-gateway
+  namespace: default
+spec:
+  # The selector matches the ingress gateway pod labels.
+  # If you installed Istio using Helm following the standard documentation, this would be "istio=ingress"
+  selector:
+    istio: ingressgateway
+  servers:
+  - port:
+      number: 80
+      name: http
+      protocol: TCP
+    hosts:
+    - ssh2.nhngpuaas.com
+---
+# virtual-service.yaml
+apiVersion: networking.istio.io/v1alpha3
+kind: VirtualService
+metadata:
+  name: test-2
+  namespace: default
+spec:
+  hosts:
+  - "ssh2.nhngpuaas.com"
+  gateways:
+  - ssh2-gateway
+  tcp:
+  - match:
+    - port: 8080
+    route:
+    - destination:
+        port:
+          number: 2222
+        host: ssh-service-2

gateway2.yaml

@@ -0,0 +1,38 @@
+# Gateway.yaml
+apiVersion: networking.istio.io/v1alpha3
+kind: Gateway
+metadata:
+  name: ssh-gateway
+  namespace: org1
+spec:
+  # The selector matches the ingress gateway pod labels.
+  # If you installed Istio using Helm following the standard documentation, this would be "istio=ingress"
+  selector:
+    istio: ingressgateway
+  servers:
+  - port:
+      number: 32222
+      name: http
+      protocol: TCP
+    hosts:
+    - ssh1.nhngpuaas.com
+---
+# virtual-service.yaml
+apiVersion: networking.istio.io/v1alpha3
+kind: VirtualService
+metadata:
+  name: test
+  namespace: org1
+spec:
+  hosts:
+  - "ssh1.nhngpuaas.com"
+  gateways:
+  - ssh-gateway
+  tcp:
+  - match:
+    - port: 32222
+    route:
+    - destination:
+        port:
+          number: 2222
+        host: ssh-service

gpuaas/nfs-provisioner/test-claim.yaml

@@ -4,7 +4,7 @@ metadata:
   name: test-claim
   namespace: istio-system
 spec:
-  storageClassName: nfs-client
+  storageClassName: user-nas
   accessModes:
     - ReadWriteMany
   resources:

gpuaas/nfs-provisioner/user-sc.yaml

@@ -0,0 +1,9 @@
+apiVersion: storage.k8s.io/v1
+kind: StorageClass
+metadata:
+  name: user-nas
+provisioner: k8s-sigs.io/nfs-subdir-external-provisioner # or choose another name, must match deployment's env PROVISIONER_NAME'
+parameters:
+  archiveOnDelete: "false"
+  onDelete: Retain
+reclaimPolicy: Retain

gpuaas/sample_pod/tf-istio.yaml

@@ -0,0 +1,110 @@
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: tf-notebook-pvc-istio
+  namespace: sample-istio
+spec:
+  accessModes:
+    - ReadWriteOnce
+  resources:
+    requests:
+      storage: 1Gi
+  storageClassName: nfs-client
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: tf-notebook-istio
+  namespace: sample-istio
+  labels:
+    app: tf-notebook-istio
+spec:
+  type: ClusterIP
+  ports:
+    - port: 80
+      name: http
+      targetPort: 8888
+  selector:
+    app: tf-notebook-istio
+---
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: tf-notebook-istio
+  namespace: sample-istio
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: tf-notebook-istio
+  template:
+    metadata:
+      labels:
+        app: tf-notebook-istio
+    spec:
+      nodeSelector:
+        nodegroup: gpu
+      containers:
+        - name: tf-notebook
+          image: tensorflow/tensorflow:2.16.2-gpu-jupyter
+          resources:
+            limits:
+              nvidia.com/gpu: 1
+          ports:
+            - containerPort: 8888
+              name: notebook
+          volumeMounts:
+            - name: notebook-storage
+              mountPath: /sample
+      volumes:
+        - name: notebook-storage
+          persistentVolumeClaim:
+            claimName: tf-notebook-pvc-istio
+---
+# Gateway.yaml
+apiVersion: networking.istio.io/v1alpha3
+kind: Gateway
+metadata:
+  name: tf-istio-gateway
+  namespace: sample-istio
+spec:
+  # The selector matches the ingress gateway pod labels.
+  # If you installed Istio using Helm following the standard documentation, this would be "istio=ingress"
+  selector:
+    istio: ingressgateway
+  servers:
+  - port:
+      number: 80
+      name: http
+      protocol: HTTP
+    hosts:
+    - tf-istio.nhngpuaas.com
+    tls:
+      httpsRedirect: true # sends 301 redirect for http requests
+  - port:
+      number: 443
+      name: https
+      protocol: HTTPS
+    hosts:
+    - tf-istio.nhngpuaas.com
+    tls:
+      mode: SIMPLE # enables HTTPS on this port
+      credentialName: nhngpuaas-ssl
+---
+# virtual-service.yaml
+apiVersion: networking.istio.io/v1alpha3
+kind: VirtualService
+metadata:
+  name: test
+  namespace: sample-istio
+spec:
+  hosts:
+  - "tf-istio.nhngpuaas.com"
+  gateways:
+  - tf-istio-gateway
+  http:
+  - route:
+    - destination:
+        port:
+          number: 80
+        host: tf-notebook-istio

monitoring/fail.yaml

@@ -0,0 +1,22 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: a1-failed-deploy
+  namespace: org1
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: fail-app
+  template:
+    metadata:
+      labels:
+        app: fail-app
+    spec:
+      nodeSelector: 
+        nodegroup: gpu
+      containers:
+      - name: fail-container
+        image: busybox
+        command: ["sh", "-c", "exit 1"]  # 실행하자마자 실패
+

monitoring/pending.yaml

@@ -0,0 +1,26 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: a1-pending-deploy
+  namespace: org1
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: pending-app
+  template:
+    metadata:
+      labels:
+        app: pending-app
+    spec:
+      nodeSelector:
+        nodegroup: warkj
+      containers:
+      - name: pending-container
+        image: busybox
+        command: ["sleep", "3600"]
+        resources:
+          requests:
+            memory: "50000Gi"  # 과도한 리소스 요청으로 스케줄링 불가
+            cpu: "100000"
+

monitoring/scp.sh

@@ -2,7 +2,7 @@
 
 # 변수 정의
 NAMESPACE="org1"       # 예: org1
-POD_NAME="a7-tf-notebook-5b5ddb5f7-mmlqv"         # 예: tf-notebook-abcdef
+POD_NAME="a7-tf-notebook-7987c576c6-d2jx8"         # 예: tf-notebook-abcdef
 LOCAL_FILE="./unlimit.py"   # 예: ./config.json
 TARGET_PATH="/tf"       # 예: /app/config/
 

svc.yaml

@@ -0,0 +1,13 @@
+apiVersion: v1
+kind: Service
+metadata:
+  name: ssh-service
+spec:
+  selector:
+    app: ssh-server
+  type: ClusterIP
+  ports:
+  - protocol: TCP
+    port: 2222
+    targetPort: 2222
+

svc2.yaml

@@ -0,0 +1,13 @@
+apiVersion: v1
+kind: Service
+metadata:
+  name: ssh-service-2
+spec:
+  selector:
+    app: ssh-server-2
+  type: ClusterIP
+  ports:
+  - protocol: TCP
+    port: 2222
+    targetPort: 2222
+

test.yaml

@@ -0,0 +1,33 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: ssh-server
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: ssh-server
+  template:
+    metadata:
+      labels:
+        app: ssh-server
+    spec:
+      containers:
+      - name: ssh-server
+        image: linuxserver/openssh-server
+        env:
+        - name: PUID
+          value: "1000"
+        - name: PGID
+          value: "1000"
+        - name: TZ
+          value: "Asia/Seoul"
+        - name: PASSWORD_ACCESS
+          value: "true"
+        - name: USER_PASSWORD
+          value: "yourpassword"   # 사용자 비밀번호 설정
+        - name: USER_NAME
+          value: "testuser"       # 사용자 이름 설정
+        ports:
+        - containerPort: 2222     # ssh는 기본적으로 22번 포트지만 이 이미지는 2222 사용
+

test2.yaml

@@ -0,0 +1,33 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: ssh-server-2
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: ssh-server-2
+  template:
+    metadata:
+      labels:
+        app: ssh-server-2
+    spec:
+      containers:
+      - name: ssh-server
+        image: linuxserver/openssh-server
+        env:
+        - name: PUID
+          value: "1000"
+        - name: PGID
+          value: "1000"
+        - name: TZ
+          value: "Asia/Seoul"
+        - name: PASSWORD_ACCESS
+          value: "true"
+        - name: USER_PASSWORD
+          value: "1111"   # 사용자 비밀번호 설정
+        - name: USER_NAME
+          value: "testuser"       # 사용자 이름 설정
+        ports:
+        - containerPort: 2222     # ssh는 기본적으로 22번 포트지만 이 이미지는 2222 사용
+