Fish Speech 1.5生产环境：Docker Compose编排双服务，日志集中采集方案

1. 项目背景与需求

Fish Speech 1.5作为新一代文本转语音模型，在实际生产环境中需要稳定可靠的服务架构。传统的单服务部署方式存在日志分散、故障排查困难、服务依赖管理复杂等问题。本文将介绍如何使用Docker Compose编排Fish Speech的双服务架构，并实现日志集中采集的完整方案。

在实际生产环境中，我们需要解决以下核心问题：

• 前端WebUI（端口7860）和后端API服务（端口7861）的协同管理
• 服务依赖关系的自动化管理
• 双服务日志的集中采集和实时监控
• 快速故障定位和性能分析能力

2. Docker Compose编排方案

2.1 服务架构设计

Fish Speech 1.5采用双服务架构，需要精心设计容器编排方案：

version: '3.8'

services:
  fish-speech-backend:
    image: fish-speech-backend:1.5
    build: ./backend
    ports:
      - "7861:7861"
    volumes:
      - ./checkpoints:/app/checkpoints
      - ./logs:/app/logs
    environment:
      - CUDA_VISIBLE_DEVICES=0
      - PYTHONPATH=/app
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]

  fish-speech-frontend:
    image: fish-speech-frontend:1.5
    build: ./frontend
    ports:
      - "7860:7860"
    depends_on:
      - fish-speech-backend
    environment:
      - BACKEND_URL=http://fish-speech-backend:7861
      - GRADIO_CDN=false

2.2 容器网络配置

为确保服务间通信安全可靠，我们配置独立的Docker网络：

networks:
  fish-speech-net:
    driver: bridge
    ipam:
      config:
        - subnet: 172.28.0.0/16

services:
  fish-speech-backend:
    networks:
      - fish-speech-net
    ports:
      - "7861:7861"

  fish-speech-frontend:
    networks:
      - fish-speech-net
    ports:
      - "7860:7860"

2.3 健康检查配置

添加健康检查确保服务稳定性：

services:
  fish-speech-backend:
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:7861/docs"]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 40s

  fish-speech-frontend:
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:7860"]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 60s

3. 日志集中采集方案

3.1 日志输出标准化

首先统一双服务的日志格式：

# 后端服务日志配置
import logging
import json
from datetime import datetime

def setup_logging():
    logging.basicConfig(
        level=logging.INFO,
        format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
        handlers=[
            logging.FileHandler('/app/logs/backend.log'),
            logging.StreamHandler()
        ]
    )

# 前端服务日志配置
def setup_frontend_logging():
    logging.basicConfig(
        level=logging.INFO,
        format='%(asctime)s - FRONTEND - %(levelname)s - %(message)s',
        handlers=[
            logging.FileHandler('/app/logs/frontend.log'),
            logging.StreamHandler()
        ]
    )

3.2 Docker日志驱动配置

使用json-file日志驱动并配置日志轮转：

services:
  fish-speech-backend:
    logging:
      driver: "json-file"
      options:
        max-size: "10m"
        max-file: "3"
        tag: "fish-speech-backend"

  fish-speech-frontend:
    logging:
      driver: "json-file"
      options:
        max-size: "10m"
        max-file: "3"
        tag: "fish-speech-frontend"

3.3 使用Fluentd实现日志收集

部署Fluentd作为日志收集器：

services:
  fluentd:
    image: fluent/fluentd:v1.16-1
    volumes:
      - ./fluentd.conf:/fluentd/etc/fluent.conf
      - ./logs:/fluentd/log
    ports:
      - "24224:24224"
      - "24224:24224/udp"
    networks:
      - fish-speech-net

  fish-speech-backend:
    logging:
      driver: "fluentd"
      options:
        tag: "fish-speech.backend"
        fluentd-address: "fluentd:24224"

  fish-speech-frontend:
    logging:
      driver: "fluentd"
      options:
        tag: "fish-speech.frontend"
        fluentd-address: "fluentd:24224"

3.4 Fluentd配置示例

创建Fluentd配置文件收集和转发日志：

<source>
  @type forward
  port 24224
  bind 0.0.0.0
</source>

<filter fish-speech.**>
  @type parser
  key_name log
  format json
  reserve_data true
</filter>

<match fish-speech.**>
  @type copy
  
  <store>
    @type file
    path /fluentd/log/fish-speech
    compress gzip
    <buffer>
      timekey 1h
      timekey_wait 10m
      timekey_use_utc true
    </buffer>
  </store>
  
  <store>
    @type elasticsearch
    host elasticsearch
    port 9200
    index_name fish-speech-%Y%m%d
    <buffer>
      timekey 1h
      timekey_wait 10m
      timekey_use_utc true
    </buffer>
  </store>
</match>

4. 完整生产环境部署

4.1 目录结构规划

fish-speech-production/
├── docker-compose.yml
├── backend/
│   ├── Dockerfile
│   └── app/
├── frontend/
│   ├── Dockerfile
│   └── app/
├── fluentd/
│   ├── Dockerfile
│   └── fluent.conf
├── elasticsearch/
│   └── config/
├── kibana/
│   └── config/
└── logs/
    ├── backend/
    ├── frontend/
    └── fluentd/

4.2 完整docker-compose配置

version: '3.8'

services:
  # Elasticsearch用于日志存储和检索
  elasticsearch:
    image: docker.elastic.co/elasticsearch/elasticsearch:8.11.0
    environment:
      - discovery.type=single-node
      - xpack.security.enabled=false
    volumes:
      - elasticsearch_data:/usr/share/elasticsearch/data
    ports:
      - "9200:9200"
    networks:
      - fish-speech-net

  # Kibana用于日志可视化
  kibana:
    image: docker.elastic.co/kibana/kibana:8.11.0
    depends_on:
      - elasticsearch
    environment:
      - ELASTICSEARCH_HOSTS=http://elasticsearch:9200
    ports:
      - "5601:5601"
    networks:
      - fish-speech-net

  # Fluentd日志收集器
  fluentd:
    image: fluent/fluentd:v1.16-1
    volumes:
      - ./fluentd/fluent.conf:/fluentd/etc/fluent.conf
      - ./logs/fluentd:/fluentd/log
    ports:
      - "24224:24224"
      - "24224:24224/udp"
    networks:
      - fish-speech-net

  # Fish Speech后端服务
  fish-speech-backend:
    build: ./backend
    ports:
      - "7861:7861"
    volumes:
      - ./checkpoints:/app/checkpoints
      - ./logs/backend:/app/logs
    environment:
      - CUDA_VISIBLE_DEVICES=0
      - PYTHONPATH=/app
    logging:
      driver: "fluentd"
      options:
        tag: "fish-speech.backend"
        fluentd-address: "fluentd:24224"
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]
    networks:
      - fish-speech-net

  # Fish Speech前端服务
  fish-speech-frontend:
    build: ./frontend
    ports:
      - "7860:7860"
    depends_on:
      - fish-speech-backend
    environment:
      - BACKEND_URL=http://fish-speech-backend:7861
      - GRADIO_CDN=false
    logging:
      driver: "fluentd"
      options:
        tag: "fish-speech.frontend"
        fluentd-address: "fluentd:24224"
    networks:
      - fish-speech-net

volumes:
  elasticsearch_data:

networks:
  fish-speech-net:
    driver: bridge

4.3 服务启动与管理

使用docker-compose管理整个服务栈：

# 启动所有服务
docker-compose up -d

# 查看服务状态
docker-compose ps

# 查看日志
docker-compose logs -f fish-speech-backend
docker-compose logs -f fish-speech-frontend

# 停止服务
docker-compose down

# 重新构建并启动
docker-compose up -d --build

5. 监控与告警配置

5.1 服务健康监控

配置Prometheus监控服务状态：

# 在docker-compose中添加监控服务
monitoring:
  image: prom/prometheus:latest
  ports:
    - "9090:9090"
  volumes:
    - ./prometheus.yml:/etc/prometheus/prometheus.yml
  networks:
    - fish-speech-net

# 配置Prometheus抓取目标
scrape_configs:
  - job_name: 'fish-speech'
    static_configs:
      - targets: ['fish-speech-backend:7861', 'fish-speech-frontend:7860']

5.2 日志告警规则

在Kibana中配置关键日志告警：

{
  "alert": {
    "name": "Fish Speech服务异常",
    "condition": {
      "script": {
        "source": "ctx.payload.hits.total.value > 0",
        "lang": "painless"
      }
    },
    "triggers": [
      {
        "name": "错误日志触发",
        "severity": "warning",
        "condition": {
          "query": {
            "match": {
              "level": "ERROR"
            }
          }
        }
      }
    ]
  }
}

6. 性能优化建议

6.1 日志性能优化

针对高并发场景优化日志性能：

# 使用异步日志记录
import logging
import asyncio
from logging.handlers import QueueHandler, QueueListener

log_queue = asyncio.Queue()
queue_handler = QueueHandler(log_queue)
listener = QueueListener(log_queue, logging.FileHandler('app.log'))
listener.start()

# 在生产代码中使用异步日志
async def process_tts_request(text):
    logger.info("开始处理TTS请求", extra={'text_length': len(text)})
    # 处理逻辑

6.2 容器资源限制

合理配置容器资源限制：

services:
  fish-speech-backend:
    deploy:
      resources:
        limits:
          cpus: '4'
          memory: 8G
        reservations:
          cpus: '2'
          memory: 6G
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]

7. 总结

通过Docker Compose编排Fish Speech 1.5的双服务架构，我们实现了以下目标：

1. 服务管理标准化：使用容器化部署，确保环境一致性和可重复性
2. 日志集中采集：通过Fluentd实现双服务日志的统一收集和存储
3. 监控可视化：集成Elasticsearch和Kibana提供强大的日志查询和分析能力
4. 高可用保障：配置健康检查和资源限制，确保服务稳定性
5. 扩展性良好：架构支持水平扩展和组件替换

这种方案特别适合生产环境部署，能够有效降低运维复杂度，提高故障排查效率，为Fish Speech 1.5的稳定运行提供有力保障。

实际部署时，建议根据具体业务需求调整资源配置和监控策略，确保系统既满足性能要求又具备良好的可维护性。

---

获取更多AI镜像

想探索更多AI镜像和应用场景？访问 CSDN星图镜像广场，提供丰富的预置镜像，覆盖大模型推理、图像生成、视频生成、模型微调等多个领域，支持一键部署。