OpenClaw 多Agent多Skill分布式部署完整方案
OpenClaw 多Agent多Skill分布式部署完整方案
从单Agent演进到多Agent协作系统，每个环节都有生产级别的细节。
🏗️ 架构概览
系统分层
┌─────────────────────────────────────────────────────────────────┐
│ 用户界面层 (Web/Mobile/API) │
└────────────────┬────────────────────────────────────┬────────────┘
│ │
┌─────────▼──────────┐ ┌──────────▼────────────┐
│ Gateway Service │ │ WebSocket Server │
│ (请求路由) │ │ (实时消息推送) │
└─────────┬──────────┘ └──────────┬────────────┘
│ │
┌─────────▼────────────────────────────────────▼──────────┐
│ Agent Management Layer (Orchestration) │
│ ┌──────────┐ ┌──────────┐ ┌──────────┐ │
│ │ Agent-A1 │ │ Agent-B2 │ │ Agent-C3 │ ... │
│ │(AI工程师)│ │(产品经理)│ │(架构师) │ │
│ └─────┬────┘ └─────┬────┘ └─────┬────┘ │
└────────┼─────────────┼─────────────┼────────────────────┘
│ │ │
┌────────▼─────────────▼─────────────▼──────────────┐
│ Skill Execution Layer (Worker Pool) │
│ ┌──────┐ ┌──────┐ ┌──────┐ ┌──────┐ ┌──────┐ │
│ │Skill │ │Skill │ │Skill │ │Skill │ │Skill │ │
│ │ 1 │ │ 2 │ │ 3 │ │ 4 │ │ N │ │
│ └──────┘ └──────┘ └──────┘ └──────┘ └──────┘ │
└────────────────────────────────────────────────────┘
│
┌────────────▼──────────────────────┐
│ Data & Service Layer │
│ ┌──────┐ ┌──────┐ ┌──────┐ │
│ │Redis │ │Kafka │ │MySQL │ ... │
│ └──────┘ └──────┘ └──────┘ │
└─────────────────────────────────────┘
核心组件
|组件|职责|部署位置|技术栈| |Gateway|HTTP请求分配、速率限制、认证|主节点|Nginx / Envoy| |Agent Orchestrator|Agent生命周期管理、任务分配|主节点|OpenClaw Core| |Skill Worker|执行具体Skill、资源隔离|Worker节点(N)|Container / Sandbox| |Message Broker|异步任务、事件驱动|独立节点|Redis / Kafka| |State Store|Agent状态持久化|独立节点|TimescaleDB / PostgreSQL| |Monitoring|系统可观测性|独立节点|Prometheus + Grafana|
组件
职责
部署位置
技术栈
Gateway
HTTP请求分配、速率限制、认证
主节点
Nginx / Envoy
Agent Orchestrator
Agent生命周期管理、任务分配
主节点
OpenClaw Core
Skill Worker
执行具体Skill、资源隔离
Worker节点(N)
Container / Sandbox
Message Broker
异步任务、事件驱动
独立节点
Redis / Kafka
State Store
Agent状态持久化
独立节点
TimescaleDB / PostgreSQL
Monitoring
系统可观测性
独立节点
Prometheus + Grafana
📁 多Agent工作区结构
目录布局
openclaw-workspace/
├── agents/ # Agent定义库
│ ├── ai-engineer/ # Agent 1: AI工程师
│ │ ├── IDENTITY.md
│ │ ├── SOUL.md
│ │ ├── config.yaml
│ │ └── tools-config.json
│ │
│ ├── product-manager/ # Agent 2: 产品经理
│ │ ├── IDENTITY.md
│ │ ├── SOUL.md
│ │ ├── config.yaml
│ │ └── tools-config.json
│ │
│ └── architect/ # Agent 3: 架构师
│ ├── IDENTITY.md
│ ├── SOUL.md
│ ├── config.yaml
│ └── tools-config.json
│
├── skills/ # 共享Skill库
│ ├── ai-engineer/
│ │ ├── SKILL.md
│ │ ├── handlers/
│ │ │ ├── model_training.py
│ │ │ ├── model_deployment.py
│ │ │ └── model_evaluation.py
│ │ └── config.yaml
│ │
│ ├── product-design/
│ │ ├── SKILL.md
│ │ ├── handlers/
│ │ └── config.yaml
│ │
│ ├── system-design/
│ │ ├── SKILL.md
│ │ ├── handlers/
│ │ └── config.yaml
│ │
│ └── shared/ # 通用Skill（所有Agent可用）
│ ├── web-search/
│ ├── file-management/
│ ├── data-analysis/
│ └── communication/
│
├── gateway/ # 网关配置
│ ├── nginx.conf
│ ├── envoy.yaml
│ └── rate-limiter.yaml
│
├── orchestrator/ # Agent编排配置
│ ├── agent-registry.yaml # Agent注册表
│ ├── skill-registry # Skill注册表
│ ├── routing-rules.yaml # 路由规则
│ └── load-balancer.yaml # 负载均衡配置
│
├── infrastructure/ # 基础设施即代码
│ ├── docker-compose.yml # 本地开发
│ ├── kubernetes/
│ │ ├── deployment.yaml
│ │ ├── service.yaml
│ │ ├── configmap.yaml
│ │ └── persistent-volumes.yaml
│ └── terraform/ # IaC定义
│ ├── main.tf
│ ├── variables.tf
│ └── outputs.tf
│
├── monitoring/ # 监控配置
│ ├── prometheus.yml
│ ├── grafana/
│ │ ├── dashboards/
│ │ └── datasources/
│ └── alerting-rules.yaml
│
├── logging/ # 日志聚合
│ ├── fluent-bit.conf
│ ├── elasticsearch/
│ └── kibana-dashboards/
│
└── docs/ # 文档
├── ARCHITECTURE.md
├── DEPLOYMENT.md
├── OPERATIONAL-GUIDE.md
└── TROUBLESHOOTING.md
🤖 Agent定义示例
1. AI工程师Agent配置
agents/ai-engineer/IDENTITY.md
# AI工程师Agent
- Name: AI工程师
- ID: agent-ai-001
- Tier: Premium
- Version: 2.1.0
- Avatar: https://...
- Emoji: 🚀
agents/ai-engineer/config.yaml
agent:
id: agent-ai-001
name: AI工程师
description: 专精ML模型开发与部署
# 模型配置
model:
provider: anthropic
name: claude-opus
version: latest
temperature: 0.3
max_tokens: 8000
thinking: moderate
# 权限配置
permissions:
- gpu_access: true
max_allocation: 2
- file_access: true
paths:
- /workspace/models
- /workspace/datasets
- network_access: true
allowed_endpoints:
- "*.huggingface.co"
- "*.nvidia.com"
# Skill绑定
skills:
- skill_id: ai-engineer
version: latest
timeout: 3600
retry_policy:
max_retries: 3
backoff: exponential
- skill_id: shared/web-search
version: latest
timeout: 30
- skill_id: shared/data-analysis
version: latest
timeout: 300
# 内存配置
memory:
type: persistent
ttl: 2592000 # 30天
storage:
backend: postgresql
connection_string: "postgresql://user:pass@db:5432/agent_ai_001"
# 速率限制
rate_limits:
requests_per_minute: 100
tokens_per_day: 1000000
concurrent_sessions: 10
# 监控配置
monitoring:
enabled: true
metrics_enabled: true
tracing_enabled: true
log_level: INFO
agents/ai-engineer/SOUL.md
# Who I Am
我是AI工程师，在模型开发和工程化落地之间架桥的实战派。
从论文到生产，每一个环节我都要把控。
## How I Talk
数据说话，不追最新论文，选最适合业务的方案。
有风险时直接预警，不藏着掖着。
## Boundaries
- 没有baseline的实验不做
- 没有离线评估的模型不上线
- 推理服务必须有降级策略
- GPU资源按需申请，用完及时释放
2. 产品经理Agent配置
agents/product-manager/config.yaml
agent:
id: agent-pm-002
name: 产品经理
description: 专精产品策划与需求分析
model:
provider: anthropic
name: claude-sonnet
version: latest
temperature: 0.5
max_tokens: 4000
permissions:
- file_access: true
paths:
- /workspace/product-specs
- /workspace/user-research
- network_access: true
allowed_endpoints:
- "*.analytics.google.com"
- "api.mixpanel.com"
skills:
- skill_id: product-design
version: latest
timeout: 1800
- skill_id: shared/data-analysis
version: latest
- skill_id: shared/web-search
version: latest
rate_limits:
requests_per_minute: 60
tokens_per_day: 500000
concurrent_sessions: 5
3. 架构师Agent配置
agents/architect/config.yaml
agent:
id: agent-arch-003
name: 架构师
description: 专精系统架构设计与技术方案
model:
provider: anthropic
name: claude-opus
version: latest
temperature: 0.2
max_tokens: 6000
permissions:
- file_access: true
paths:
- /workspace/architecture
- /workspace/infrastructure
- network_access: true
allowed_endpoints:
- "*.docker.com"
- "*.kubernetes.io"
skills:
- skill_id: system-design
version: latest
timeout: 2400
- skill_id: shared/web-search
version: latest
- skill_id: shared/file-management
version: latest
rate_limits:
requests_per_minute: 80
tokens_per_day: 800000
concurrent_sessions: 8
🛠️ Skill定义与共享
Skill注册表
orchestrator/skill-registry.yaml
skills:
# Agent专用Skill
- id: ai-engineer
name: AI工程师能力包
version: 2.1.0
location: /workspace/skills/ai-engineer
owner: ai-engineer
visibility: private # 仅该Agent可用
resource_requirements:
cpu: 2
memory: 4Gi
gpu: optional
timeout: 3600
handlers:
- model_training
- model_deployment
- model_evaluation
- id: product-design
name: 产品设计能力包
version: 1.5.0
location: /workspace/skills/product-design
owner: product-manager
visibility: private
resource_requirements:
cpu: 1
memory: 2Gi
gpu: none
timeout: 1800
- id: system-design
name: 系统架构能力包
version: 1.2.0
location: /workspace/skills/system-design
owner: architect
visibility: private
resource_requirements:
cpu: 1
memory: 2Gi
timeout: 2400
# 共享Skill（所有Agent可用）
- id: shared/web-search
name: 网页搜索
version: 3.0.0
location: /workspace/skills/shared/web-search
owner: system
visibility: public
resource_requirements:
cpu: 1
memory: 1Gi
timeout: 30
rate_limit: 100/minute
- id: shared/data-analysis
name: 数据分析
version: 2.1.0
location: /workspace/skills/shared/data-analysis
owner: system
visibility: public
resource_requirements:
cpu: 2
memory: 4Gi
gpu: optional
timeout: 300
- id: shared/file-management
name: 文件管理
version: 1.8.0
location: /workspace/skills/shared/file-management
owner: system
visibility: public
resource_requirements:
cpu: 1
memory: 1Gi
timeout: 60
- id: shared/communication
name: 沟通协作
version: 1.3.0
location: /workspace/skills/shared/communication
owner: system
visibility: public
resource_requirements:
cpu: 1
memory: 1Gi
timeout: 30
🔀 Agent编排与路由
Agent注册表
orchestrator/agent-registry.yaml
agents:
- id: agent-ai-001
name: AI工程师
status: active
version: 2.1.0
replicas: 3 # 3个并发副本
# 专业领域
domains:
- machine_learning
- model_deployment
- data_engineering
# 服务质量承诺
slo:
availability: 99.9%
response_time_p99: 5000ms
throughput: 100req/min
# 资源分配
resources:
requests:
cpu: 2
memory: 4Gi
limits:
cpu: 4
memory: 8Gi
# 扩缩容策略
autoscaling:
enabled: true
min_replicas: 1
max_replicas: 10
target_cpu_utilization: 70%
target_memory_utilization: 75%
- id: agent-pm-002
name: 产品经理
status: active
version: 1.2.0
replicas: 2
domains:
- product_strategy
- user_research
- requirements_analysis
slo:
availability: 99.5%
response_time_p99: 3000ms
throughput: 60req/min
resources:
requests:
cpu: 1
memory: 2Gi
limits:
cpu: 2
memory: 4Gi
autoscaling:
enabled: true
min_replicas: 1
max_replicas: 5
- id: agent-arch-003
name: 架构师
status: active
version: 1.1.0
replicas: 2
domains:
- system_architecture
- infrastructure_design
- technology_selection
slo:
availability: 99.5%
response_time_p99: 4000ms
throughput: 50req/min
resources:
requests:
cpu: 1
memory: 2Gi
limits:
cpu: 2
memory: 4Gi
路由规则
orchestrator/routing-rules.yaml
routing_rules:
# 基于意图的路由
- name: ml_model_training
pattern: "(训练|模型|深度学习|神经网络)"
agents:
- agent-ai-001
priority: 10
timeout: 3600
- name: product_strategy
pattern: "(产品|需求|用户|功能)"
agents:
- agent-pm-002
priority: 8
timeout: 1800
- name: system_architecture
pattern: "(架构|设计|系统|基础设施)"
agents:
- agent-arch-003
priority: 8
timeout: 2400
# 多Agent协作流程
- name: end_to_end_feature_delivery
pattern: "(端到端|完整方案|从需求到上线)"
workflow:
- agent: agent-pm-002
step: 1
task: 分析需求和用户研究
timeout: 1800
- agent: agent-arch-003
step: 2
task: 制定技术架构方案
depends_on: step-1
timeout: 2400
- agent: agent-ai-001
step: 3
task: 实现和部署模型
depends_on: step-2
timeout: 3600
priority: 15
timeout: 7800
# 默认路由（通用Skill）
- name: default_shared_skills
pattern: ".*"
agents: null # 不指定Agent，直接执行shared skill
priority: 1
timeout: 300
🐳 部署配置
Docker Compose（本地开发）
infrastructure/docker-compose.yml
version: '3.9'
services:
# 网关
gateway:
image: nginx:latest
ports:
- "8080:80"
volumes:
- ./gateway/nginx.conf:/etc/nginx/nginx.conf:ro
depends_on:
- orchestrator
networks:
- openclaw
# Agent编排器
orchestrator:
build:
context: .
dockerfile: Dockerfile.orchestrator
environment:
- AGENT_REGISTRY_PATH=/workspace/orchestrator/agent-registry.yaml
- SKILL_REGISTRY_PATH=/workspace/orchestrator/skill-registry.yaml
- REDIS_URL=redis://redis:6379
- DATABASE_URL=postgresql://user:pass@postgres:5432/openclaw
- LOG_LEVEL=INFO
ports:
- "8888:8888"
volumes:
- ./orchestrator:/workspace/orchestrator:ro
- ./agents:/workspace/agents:ro
- ./skills:/workspace/skills:ro
depends_on:
- redis
- postgres
- kafka
networks:
- openclaw
# Agent Worker 1
agent-ai-worker-1:
build:
context: .
dockerfile: Dockerfile.agent
environment:
- AGENT_ID=agent-ai-001
- WORKER_ID=worker-ai-1
- ORCHESTRATOR_URL=http://orchestrator:8888
- REDIS_URL=redis://redis:6379
- SKILL_PATHS=/workspace/skills
volumes:
- ./agents:/workspace/agents:ro
- ./skills:/workspace/skills:ro
- agent-ai-1-cache:/cache
depends_on:
- orchestrator
networks:
- openclaw
# Agent Worker 2
agent-pm-worker-1:
build:
context: .
dockerfile: Dockerfile.agent
environment:
- AGENT_ID=agent-pm-002
- WORKER_ID=worker-pm-1
- ORCHESTRATOR_URL=http://orchestrator:8888
- REDIS_URL=redis://redis:6379
volumes:
- ./agents:/workspace/agents:ro
- ./skills:/workspace/skills:ro
- agent-pm-1-cache:/cache
depends_on:
- orchestrator
networks:
- openclaw
# Agent Worker 3
agent-arch-worker-1:
build:
context: .
dockerfile: Dockerfile.agent
environment:
- AGENT_ID=agent-arch-003
- WORKER_ID=worker-arch-1
- ORCHESTRATOR_URL=http://orchestrator:8888
- REDIS_URL=redis://redis:6379
volumes:
- ./agents:/workspace/agents:ro
- ./skills:/workspace/skills:ro
- agent-arch-1-cache:/cache
depends_on:
- orchestrator
networks:
- openclaw
# 消息队列
redis:
image: redis:7-alpine
ports:
- "6379:6379"
volumes:
- redis-data:/data
networks:
- openclaw
kafka:
image: confluentinc/cp-kafka:7.4.0
environment:
KAFKA_BROKER_ID: 1
KAFKA_ZOOKEEPER_CONNECT: zookeeper:2181
KAFKA_ADVERTISED_LISTENERS: PLAINTEXT://kafka:9092
KAFKA_OFFSETS_TOPIC_REPLICATION_FACTOR: 1
depends_on:
- zookeeper
networks:
- openclaw
zookeeper:
image: confluentinc/cp-zookeeper:7.4.0
environment:
ZOOKEEPER_CLIENT_PORT: 2181
networks:
- openclaw
# 数据库
postgres:
image: postgres:15-alpine
environment:
POSTGRES_USER: user
POSTGRES_PASSWORD: pass
POSTGRES_DB: openclaw
volumes:
- postgres-data:/var/lib/postgresql/data
networks:
- openclaw
# 监控
prometheus:
image: prom/prometheus:latest
volumes:
- ./monitoring/prometheus.yml:/etc/prometheus/prometheus.yml:ro
- prometheus-data:/prometheus
ports:
- "9090:9090"
networks:
- openclaw
grafana:
image: grafana/grafana:latest
environment:
- GF_SECURITY_ADMIN_PASSWORD=admin
ports:
- "3000:3000"
volumes:
- grafana-data:/var/lib/grafana
depends_on:
- prometheus
networks:
- openclaw
volumes:
redis-data:
postgres-data:
prometheus-data:
grafana-data:
agent-ai-1-cache:
agent-pm-1-cache:
agent-arch-1-cache:
networks:
openclaw:
driver: bridge
Kubernetes部署
infrastructure/kubernetes/deployment.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
name: openclaw-gateway
namespace: openclaw
spec:
replicas: 3
selector:
matchLabels:
app: gateway
template:
metadata:
labels:
app: gateway
spec:
containers:
- name: gateway
image: openclaw/gateway:v1.0.0
ports:
- containerPort: 80
resources:
requests:
cpu: 500m
memory: 512Mi
limits:
cpu: 1000m
memory: 1Gi
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: ohestrator
namespace: openclaw
spec:
replicas: 1
selector:
matchLabels:
app: orchestrator
template:
metadata:
labels:
app: orchestrator
spec:
containers:
- name: orchestrator
image: openclaw/orchestrator:v1.0.0
ports:
- containerPort: 8888
env:
- name: AGENT_REGISTRY_PATH
value: /etc/openclaw/agent-registry.yaml
- name: SKILL_REGISTRY_PATH
value: /etc/openclaw/skill-registry.yaml
- name: REDIS_URL
value: redis://redis-service:6379
- name: DATABASE_URL
valueFrom:
secretKeyRef:
name: openclaw-secrets
key: database-url
volumeMounts:
- name: agent-registry
mountPath: /etc/openclaw/agent-registry.yaml
subPath: agent-registry.yaml
- name: skill-registry
mountPath: /etc/openclaw/skill-registry.yaml
subPath: skill-registry.yaml
resources:
requests:
cpu: 1
memory: 2Gi
limits:
cpu: 2
memory: 4Gi
volumes:
- name: agent-registry
configMap:
name: agent-registry
- name: skill-registry
configMap:
name: skill-registry
---
apiVersion: apps/v1
kind: StatefulSet
metadata:
name: openclaw-agent-ai
namespace: openclaw
spec:
serviceName: agent-ai-service
replicas: 3
selector:
matchLabels:
agent: ai-engineer
template:
metadata:
labels:
agent: ai-engineer
spec:
containers:
- name: agent-worker
image: openclaw/agent-worker:v1.0.0
env:
- name: AGENT_ID
value: agent-ai-001
- name: WORKER_ID
valueFrom:
fieldRef:
fieldPath: metadata.name
- name: ORCHESTRATOR_URL
value: http://orchestrator-service:8888
- name: REDIS_URL
value: redis://redis-service:6379
resources:
requests:
cpu: 2
memory: 4Gi
nvidia.com/gpu: 1
limits:
cpu: 4
memory: 8Gi
nvidia.com/gpu: 1
volumeMounts:
- name: cache
mountPath: /cache
- name: models
mountPath: /models
volumeClaimTemplates:
- metadata:
name: cache
spec:
accessModes: ["ReadWriteOnce"]
storageClassName: fast-ssd
resources:
requests:
storage: 100Gi
- metadata:
name: models
spec:
accessModes: ["ReadWriteOnce"]
storageClassName: nfs
resources:
requests:
storan---
apiVersion: v1
kind: Service
metadata:
name: gateway-service
namespace: openclaw
spec:
type: LoadBalancer
selector:
app: gateway
ports:
- protocol: TCP
port: 80
targetPort: 80
---
apiVersion: v1
kind: Service
metadata:
name: orchestrator-service
namespace: openclaw
spec:
selector:
app: orchestrator
ports:
- protocol: TCP
port: 8888
targetPort: 8888
📊 监控与可观测性
Prometheus配置
monitoring/prometheus.yml
global:
scrape_interval: 15s
evaluation_interval: 15s
scrape_configs:
# Agent指标
- job_name: 'agents'
static_configs:
- targets:
- 'orchestrator:8888'
- 'agent-ai-001:9100'
- 'agent-pm-002:9100'
- 'agent-arch-003:9100'
metrics_path: '/metrics'
# 基础设施指标
- job_name: 'infrastructure'
static_configs:
- targets: ['node-exporter:9100']
# Redis指标
- job_name: 'redis'
static_configs:
- targets: ['redis-exporter:9121']
alert:
alertmanagers:
- static_configs:
- targets: ['alertmanager:9093']
rule_files:
- '/etc/prometheus/rules/*.yml'
关键指标
# 监控指标定义
metrics = {
"agent_request_latency_seconds": {
"type": "histogram",
"help": "Agent请求延迟",
"labels": ["agent_id", "skill_id"],
"buckets": [0.1, 0.5, 1, 2, 5, 10]
},
"agent_requests_total": {
"type": "counter",
"help": "总请求数",
"labels": ["agent_id", "status"]
},
"agent_tokens_consumed": {
"type": "counter",
"help": "消耗的Token数",
"labels": ["agent_id"]
},
"skill_execution_duration_seconds": {
"type": "histogram",
"help": "Skill执行时间",
"labels": ["skill_id", "status"]
},
"agent_active_sessions": {
"type": "gauge",
"help": "当前活跃会话数",
"labels": ["agent_id"]
},
"message_queue_depth": {
"type": "gauge",
"help": "消息队列深度",
"labels": ["queue_name"]
}
}
🔄 多Agent协作示例
场景：端到端产品交付
workflow: end-to-end-delivery
steps:
- id: step-1
name: 需求分析
agent: agent-pm-002
inputs:
edback
- market_research
skills:
- product-design
- shared/data-analysis
expected_output: product_spec
timeout: 1800
- id: step-2
name: 技术方案
agent: agent-arch-003
inputs:
- product_spec # 来自step-1
skills:
- system-design
- shared/web-search
expected_output: tech_design_doc
timeout: 2400
depends_on: step-1
- id: step-3
name: 模型训练与部署
agent: agent-ai-001
inputs:
- tech_design_doc # 来自step-2
- training_data
skills:
- ai-engineer
- shared/data-analysis
expected_output: deployed_model
timeout: 3600
depends_on: step-2
- id: step-4
name: 交叉评审
agents:
- agent-pm-002
- agent-arch-003
- agent-ai-001
inputs:
- deployed_model
- product_spec
- tech_design_doc
skills:
- shared/communication
expected_output: review_report
timeout: 900
depends_on: step-3
total_timeout: 8700 # 2.4小时
⚙️ 运维操作
快速启动
# 本地开发环境
docker-compose up -d
# Kubernetes环境
kubectl create namespace openclaw
kubectl apply -f infrastructure/kubernetes/
# 检查所有Agent状态
kubectl get pods -n openclaw -l app=agent-worker
# 查看编排器日志
kubectl logs -n openclaw deployment/openclaw-orchestrator -f
添加新Agent
# 1. 创建Agent配置
mkdir -p agents/new-agent
cp agents/template/* agents/new-agent/
# 编辑 IDENTITY.md, SOUL.md, config.yaml
# 2. 更新Agent注册表
# 编辑 orchestrator/agent-registry.yaml
# 3. 绑定Skills
# 编辑 agents/new-agent/config.yaml 的 skills 部分
# 4. 重新部署
docker-compose restart orchestrator
# 或
kubectl rollout restart deployment/openclaw-orchestrator -n openclaw
扩容/缩容
# Kubernetes自动扩缩容
kubectl autoscale statefulset openclaw-agent-ai \
--min=1 --max=10 --cpu-percent=70 -n openclaw
# 手动扩容到N个副本
kubectl scale statefulset openclaw-agent-ai --replicas=5 -n openclaw
✅ 部署检查清单
- [ ] 所有Agent配置文件已创建（IDENTITY.md、SOUL.md、config.yaml）
- [ ] 所有Skill已注册到skill-registry.yaml
- [ ] 路由规则已配置到routing-rules.yaml
- [ ] 网关已配置（Nginx/Envoy）
- [ ] 消息队列（Redis/Kafka）已部署
- [ ] 数据库已迁移
- [ ] 监控告警已配置
- [ ] 日志聚合已启用
- [ ] 本地docker-compose测试通过
- [ ] Kubernetes集群已准备就绪
- [ ] 多Agent协作流程已测试
- [ ] 负载均衡已验证
- [ ] 故障转移已验证
- [ ] 文档已更新
评论 (0)
发表评论