1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23

CLUSTER_NAME=my-cluster
CLUSTER_LOCATION=us-central1

gcloud container clusters create-auto ${CLUSTER_NAME} \
  --location=${CLUSTER_LOCATION}

# kubeconfig entry generated for my-cluster.
# NAME: my-cluster
# LOCATION: us-central1
# MASTER_VERSION: 1.30.5-gke.1014001
# MASTER_IP: 34.45.80.180
# MACHINE_TYPE: e2-small
# NODE_VERSION: 1.30.5-gke.1014001
# NUM_NODES: 3
# STATUS: RUNNING

kubectl cluster-info
# Kubernetes control plane is running at https://34.45.80.180
# GLBCDefaultBackend is running at https://34.45.80.180/api/v1/namespaces/kube-system/services/default-http-backend:http/proxy
# KubeDNS is running at https://34.45.80.180/api/v1/namespaces/kube-system/services/kube-dns:dns/proxy
# Metrics-server is running at https://34.45.80.180/api/v1/namespaces/kube-system/services/https:metrics-server:/proxy

# To further debug and diagnose cluster problems, use 'kubectl cluster-info dump'.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12

kubectl apply -f ollama.yaml
# Warning: autopilot-default-resources-mutator:Autopilot updated StatefulSet default/ollama: defaulted unspecified 'nvidia.com/gpu' resource for containers [ollama] (see http://g.co/gke/autopilot-defaults).
# statefulset.apps/ollama created
# service/ollama created

kubectl get pods -w
# 等待 ollama-0 Pod 變成 Running 狀態，表示 Ollama 已成功部署。

kubectl get nodes
# NAME                                        STATUS   ROLES    AGE     VERSION
# gk3-my-cluster-nap-1l47y8kj-033ca61f-rgvm   Ready    <none>   2m54s   v1.30.5-gke.1014001
# gk3-my-cluster-nap-nwlb0qz1-ea6bcb2c-b44d   Ready    <none>   2m9s    v1.30.5-gke.1014001

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12

# 安裝 Ollama CLI：在 Cloud Shell 中執行以下指令安裝 Ollama CLI：
curl -fsSL https://ollama.com/install.sh | sh

# 將 Ollama 服務端口轉發到本地： 執行以下指令，將 Ollama 服務的 11434 端口轉發到本地 (Cloud Shell) 的 11434 端口：
kubectl port-forward svc/ollama 11434:11434

# 在新分頁中執行 Gemma 2 模型： 在新的終端分頁中，設定 OLLAMA_HOST 環境變數，指定 Ollama API 的地址，然後使用 ollama run 指令執行 Gemma 2 模型：
OLLAMA_HOST=http://localhost:11434 ollama run gemma2

>>> 你好！

# 你好！👋  有什么我可以帮你的吗？😊

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13

student_03_216a983cfb25@cloudshell:~ (qwiklabs-asl-02-cd54d6e804de)$ 

gcloud auth list
# Credentialed Accounts

# ACTIVE: *
# ACCOUNT: [email protected]

gcloud config list project
# [core]
# project = qwiklabs-asl-02-cd54d6e804de

# Your active configuration is: [cloudshell-23006]

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23

export CLUSTER_NAME="ml-gke"
export REGION="us-central1"
export BUCKET_NAME=${GOOGLE_CLOUD_PROJECT}-llama-l4
export SERVICE_ACCOUNT="l4-lab@${GOOGLE_CLOUD_PROJECT}.iam.gserviceaccount.com"

gcloud container clusters create $CLUSTER_NAME \
  --enable-image-streaming \
  --addons=HttpLoadBalancing \
  --machine-type=e2-standard-2 \
  --shielded-secure-boot \
  --shielded-integrity-monitoring \
  --region=${REGION} \
  --num-nodes=1 \
  --enable-ip-alias \
  --release-channel=rapid \
  --node-locations=${REGION}-a \
  --workload-pool=${GOOGLE_CLOUD_PROJECT}.svc.id.goog \
  --addons GcsFuseCsiDriver

# Note: The Kubelet readonly port (10255) is now deprecated. Please update your workloads to use the recommended alternatives. See https://cloud.google.com/kubernetes-engine/docs/how-to/disable-kubelet-readonly-port for ways to check usage and for migration instructions.
# Creating cluster ml-gke in us-central1... Cluster is being health-checked (Kubernetes Control Plane is healthy)...done.                           
# Created [https://container.googleapis.com/v1/projects/qwiklabs-asl-02-cd54d6e804de/zones/us-central1/clusters/ml-gke].
# To inspect the contents of your cluster, go to: https://console.cloud.google.com/kubernetes/workload_/gcloud/us-central1/ml-gke?project=qwiklabs-asl-02-cd54d6e804de

 1
 2
 3
 4
 5
 6
 7
 8
 9
10

gcloud container clusters list

NAME: ml-gke
LOCATION: us-central1
MASTER_VERSION: 1.31.1-gke.1146000
MASTER_IP: 34.67.21.231
MACHINE_TYPE: e2-standard-2
NODE_VERSION: 1.31.1-gke.1146000
NUM_NODES: 1
STATUS: RUNNING

1
2

2024-10-23-k8s-summit-gcp-1
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

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29

# Set the default environment variables:
export HF_TOKEN=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

# Create the Node Pool
gcloud container node-pools create gpupool --cluster ml-gke \
  --accelerator type=nvidia-l4,count=8,gpu-driver-version=latest \
  --machine-type g2-standard-96 \
  --ephemeral-storage-local-ssd=count=8 \
  --enable-autoscaling --enable-image-streaming \
  --num-nodes=0 --min-nodes=0 --max-nodes=3 \
  --shielded-secure-boot \
  --shielded-integrity-monitoring \
  --node-locations ${REGION}-a,${REGION}-b --region ${REGION} 

# Note: Machines with GPUs have certain limitations which may affect your workflow. Learn more at https://cloud.google.com/kubernetes-engine/docs/how-to/gpus
# Note: Starting in GKE 1.30.1-gke.115600, if you don't specify a driver version, GKE installs the default GPU driver for your node's GKE version.
# Creating node pool gpupool...done.                                                                                                                
# Created [https://container.googleapis.com/v1/projects/qwiklabs-asl-02-cd54d6e804de/zones/us-central1/clusters/ml-gke/nodePools/gpupool].

# NAME: gpupool
# MACHINE_TYPE: g2-standard-96
# DISK_SIZE_GB: 100
# NODE_VERSION: 1.31.1-gke.1146000

# Create a Kubernetes Secret that contains the Hugging Face token
kubectl create secret generic hf-secret \
  --from-literal=hf_api_token=${HF_TOKEN} \
  --dry-run=client -o yaml | kubectl apply -f -
# secret/hf-secret created