gcp-compute

GCP Compute Engine

在 Google Cloud Platform 上部署、管理和扩展 Compute Engine 虚拟机。

适用场景

• 在 GCP 上部署 Web 服务器、应用后端或批处理工作负载
• 运行需要完全控制操作系统级别的工作负载（不同于 Cloud Run 或 App Engine）
• 创建托管实例组，在负载均衡器后实现自动修复和自动扩缩
• 配置附加 GPU 的虚拟机，用于机器学习训练或渲染流水线
• 使用可抢占式或 Spot 虚拟机，为非关键工作负载优化成本

前提条件

• 已安装并完成身份验证的 Google Cloud SDK（gcloud）
• 已启用 Compute Engine API 的 GCP 项目
• IAM 角色 roles/compute.admin 或用于实例管理的范围限定角色

gcloud auth list
gcloud config set project $PROJECT_ID
gcloud services enable compute.googleapis.com

机器类型参考

系列	示例	vCPU	内存	用途
E2	e2-micro	0.25	1 GB	开发/测试、微服务
E2	e2-medium	1	4 GB	轻量级 Web 服务器
N2	n2-standard-4	4	16 GB	通用生产环境
N2	n2-highmem-8	8	64 GB	内存缓存、数据库
C2	c2-standard-16	16	64 GB	计算密集型工作负载、高性能计算

# List machine types available in a zone
gcloud compute machine-types list --zones=us-central1-a --filter="name~'e2-'"

# Create a custom machine type (6 vCPUs, 24 GB RAM)
gcloud compute instances create custom-vm \
  --custom-cpu=6 --custom-memory=24GB \
  --zone=us-central1-a \
  --image-family=debian-12 --image-project=debian-cloud

创建实例

# Production instance with shielded VM and startup script
gcloud compute instances create web-server \
  --machine-type=e2-medium \
  --zone=us-central1-a \
  --image-family=debian-12 \
  --image-project=debian-cloud \
  --boot-disk-size=20GB \
  --boot-disk-type=pd-balanced \
  --tags=http-server,https-server \
  --labels=env=production,team=backend \
  --metadata=enable-oslogin=TRUE \
  --shielded-secure-boot \
  --shielded-vtpm \
  --shielded-integrity-monitoring

# Instance with a startup script and service account
gcloud compute instances create app-server \
  --machine-type=e2-standard-2 \
  --zone=us-central1-a \
  --image-family=ubuntu-2204-lts \
  --image-project=ubuntu-os-cloud \
  --boot-disk-size=50GB \
  --metadata-from-file=startup-script=startup.sh \
  --service-account=app-sa@${PROJECT_ID}.iam.gserviceaccount.com \
  --scopes=cloud-platform

# Instance with an additional data disk
gcloud compute instances create db-server \
  --machine-type=n2-highmem-4 \
  --zone=us-central1-a \
  --image-family=debian-12 --image-project=debian-cloud \
  --boot-disk-size=20GB \
  --create-disk=name=data-disk,size=200GB,type=pd-ssd,auto-delete=no

启动脚本示例

#!/bin/bash
# startup.sh - runs on first boot and every reboot
set -euo pipefail
apt-get update && apt-get install -y nginx
systemctl enable nginx && systemctl start nginx
curl -X PUT -H "Metadata-Flavor: Google" \
  "http://metadata.google.internal/computeMetadata/v1/instance/guest-attributes/startup/status" \
  -d "complete"

实例模板和托管实例组

# Create an instance template
gcloud compute instance-templates create web-template \
  --machine-type=e2-medium \
  --image-family=debian-12 --image-project=debian-cloud \
  --boot-disk-size=20GB --tags=http-server \
  --metadata-from-file=startup-script=startup.sh

# Create a regional managed instance group (MIG) with health check
gcloud compute health-checks create http http-health-check \
  --port=80 --request-path=/healthz \
  --check-interval=10s --timeout=5s \
  --healthy-threshold=2 --unhealthy-threshold=3

gcloud compute instance-groups managed create web-mig \
  --template=web-template --size=3 \
  --region=us-central1 \
  --health-check=http-health-check --initial-delay=120

# Configure autoscaling
gcloud compute instance-groups managed set-autoscaling web-mig \
  --region=us-central1 \
  --min-num-replicas=2 --max-num-replicas=10 \
  --target-cpu-utilization=0.65 --cool-down-period=90

# Rolling update to a new template
gcloud compute instance-groups managed rolling-action start-update web-mig \
  --version=template=web-template-v2 \
  --region=us-central1 --max-surge=3 --max-unavailable=0

可抢占式和 Spot 虚拟机

# Spot VM (recommended over legacy preemptible)
gcloud compute instances create spot-worker \
  --machine-type=n2-standard-8 \
  --zone=us-central1-a \
  --image-family=debian-12 --image-project=debian-cloud \
  --provisioning-model=SPOT \
  --instance-termination-action=STOP

# Spot instance template for batch MIG
gcloud compute instance-templates create batch-template \
  --machine-type=n2-standard-4 \
  --image-family=debian-12 --image-project=debian-cloud \
  --provisioning-model=SPOT \
  --instance-termination-action=DELETE

快照和映像

# Create a snapshot
gcloud compute disks snapshot web-server \
  --zone=us-central1-a \
  --snapshot-names=web-server-snap-$(date +%Y%m%d)

# Scheduled snapshot policy
gcloud compute resource-policies create snapshot-schedule daily-backup \
  --region=us-central1 --max-retention-days=14 \
  --daily-schedule --start-time=03:00

gcloud compute disks add-resource-policies web-server \
  --zone=us-central1-a --resource-policies=daily-backup

# Create a custom image from an instance
gcloud compute instances stop web-server --zone=us-central1-a
gcloud compute images create web-golden-image \
  --source-disk=web-server --source-disk-zone=us-central1-a \
  --family=web-server --labels=version=v1

Terraform 配置

resource "google_compute_instance" "web" {
  name         = "web-server"
  machine_type = "e2-medium"
  zone         = "us-central1-a"
  tags         = ["http-server", "https-server"]

  boot_disk {
    initialize_params {
      image = "debian-cloud/debian-12"
      size  = 20
      type  = "pd-balanced"
    }
  }

  network_interface {
    subnetwork = google_compute_subnetwork.main.id
    access_config {}
  }

  metadata_startup_script = file("${path.module}/startup.sh")

  service_account {
    email  = google_service_account.app.email
    scopes = ["cloud-platform"]
  }

  shielded_instance_config {
    enable_secure_boot          = true
    enable_vtpm                 = true
    enable_integrity_monitoring = true
  }
}

resource "google_compute_instance_template" "web" {
  name_prefix  = "web-"
  machine_type = "e2-medium"

  disk {
    source_image = "debian-cloud/debian-12"
    auto_delete  = true
    boot         = true
    disk_size_gb = 20
  }

  network_interface {
    subnetwork = google_compute_subnetwork.main.id
  }

  lifecycle { create_before_destroy = true }
}

resource "google_compute_region_instance_group_manager" "web" {
  name               = "web-mig"
  base_instance_name = "web"
  region             = "us-central1"

  version {
    instance_template = google_compute_instance_template.web.id
  }

  target_size = 3
  named_port { name = "http"; port = 80 }

  auto_healing_policies {
    health_check      = google_compute_health_check.http.id
    initial_delay_sec = 120
  }
}

resource "google_compute_region_autoscaler" "web" {
  name   = "web-autoscaler"
  region = "us-central1"
  target = google_compute_region_instance_group_manager.web.id

  autoscaling_policy {
    min_replicas    = 2
    max_replicas    = 10
    cooldown_period = 90
    cpu_utilization { target = 0.65 }
  }
}

常见操作

# SSH into an instance
gcloud compute ssh web-server --zone=us-central1-a

# List all instances with status
gcloud compute instances list \
  --format="table(name,zone,status,machineType.basename())"

# Stop / start / resize
gcloud compute instances stop web-server --zone=us-central1-a
gcloud compute instances set-machine-type web-server \
  --machine-type=e2-standard-4 --zone=us-central1-a
gcloud compute instances start web-server --zone=us-central1-a

# View serial port output (debug startup scripts)
gcloud compute instances get-serial-port-output web-server --zone=us-central1-a

故障排除

症状	原因	解决方法
实例卡在 STAGING 状态	配额超限或资源不可用	使用 gcloud compute project-info describe 检查配额；尝试其他可用区
启动脚本未运行	语法错误或元数据键错误	检查串行端口输出；确保键为 startup-script，而不是 startup_script
无法通过 SSH 连接	防火墙阻止了 22 端口，或 OS Login 配置错误	添加 tcp:22 防火墙规则；检查 enable-oslogin 元数据
过于频繁地被抢占	可用区资源紧张	使用采用 STOP 操作的 Spot 虚拟机；在 MIG 中分散到多个可用区
磁盘空间不足	启动磁盘太小	使用 gcloud compute disks resize；为数据磁盘启用 --storage-auto-increase
MIG 未自动修复	健康检查配置错误或初始延迟太短	确认健康检查路径返回 200；增大 --initial-delay

相关技能

• gcp-networking - Compute Engine 的 VPC、防火墙规则和负载均衡器
• terraform-gcp - 使用基础设施即代码配置 Compute Engine 资源
• gcp-gke - 适用于更适合使用容器而非虚拟机的工作负载
• gcp-cloud-sql - Compute Engine 应用连接的托管数据库