在使用 docker-compose 部署 Harbor 时遇到一个现象：服务有时候能正常启动，有时候启动失败。
问题日志：
ddb266c70e76 cr.loongnix.cn/harbor/harbor-jobservice:2.2.1 "/harbor/entrypoint.…" 48 seconds ago Up 38 seconds (health: starting) harbor-jobservice
89109fe1dcb0 cr.loongnix.cn/harbor/nginx-photon:2.2.1 "nginx -g 'daemon of…" 48 seconds ago Restarting (1) 13 seconds ago nginx
b965a3ab3f9c cr.loongnix.cn/harbor/harbor-core:2.2.1 "/harbor/entrypoint.…" 50 seconds ago Restarting (1) 1 second ago harbor-core
7c3a830b2821 cr.loongnix.cn/harbor/harbor-db:2.2.1 "/docker-entrypoint.…" 52 seconds ago Up 43 seconds (healthy) harbor-db
59b80ec6e264 cr.loongnix.cn/harbor/harbor-portal:2.2.1 "nginx -g 'daemon of…" 52 seconds ago Up 43 seconds (healthy) harbor-portal
5eed18a72884 cr.loongnix.cn/harbor/harbor-registryctl:2.2.1 "/home/harbor/start.…" 52 seconds ago Up 43 seconds (healthy) registryctl
8292f0b126d8 cr.loongnix.cn/harbor/registry-photon:2.2.1 "/home/harbor/entryp…" 52 seconds ago Up 43 seconds (healthy) registry
8d75abeadfda cr.loongnix.cn/harbor/redis-photon:2.2.1 "redis-server /etc/r…" 52 seconds ago Up 41 seconds (healthy) redis
09d15838f356 cr.loongnix.cn/harbor/harbor-log:2.2.1 "/bin/sh -c /usr/loc…" 53 seconds ago Up 45 seconds (healthy) 127.0.0.1:1514->10514/tcp harbor-log
通过分析启动过程发现，该问题与 Docker 容器的健康检查机制以及 docker-compose 的 depends_on依赖管理机制有关。基于此现象，我们提出以下几个具体问题：
docker-compose up -d命令时，为什么会出现depends_on依赖关系看似未生效的情况？容器的健康检查（Health Check）是用来检测容器内运行的应用是否处于健康状态的一种机制。它可以帮助你自动发现和处理异常状态的服务，确保系统更加稳定和可靠。
容器可能仍在运行，但里面的应用已经崩溃或卡死了。健康检查可以：
Docker 的 HEALTHCHECK 指令允许你指定一个命令，用于检测容器内部应用是否健康运行。当命令执行成功（退出码为 0）时，容器状态为 healthy，否则为 unhealthy。如果要为 Docker 镜像添加健康检查，则需要在 Dockerfile 中使用 HEALTHCHECK 指令:
FROM cr.loongnix.cn/library/python:3.10.14-slim-buster
WORKDIR /app
RUN pip install --no-cache-dir flask
COPY app.py .
# 添加健康检查：尝试访问 /health
HEALTHCHECK --interval=10s --timeout=3s --start-period=5s --retries=3 \
CMD curl -f http://localhost:5000/health || exit 1
CMD ["python", "app.py"]
- app.py
import time
from flask import Flask, jsonify
app = Flask(__name__)
@app.route("/")
def index():
return "Hello, World!"
@app.route("/health")
def health():
return jsonify(status="ok"), 200
if __name__ == "__main__":
time.sleep(20)
app.run(host="0.0.0.0", port=5000)
--interval=10s：Docker 每隔 10 秒 执行一次健康检查命令。--timeout=3s: 如果 curl 命令在 3 秒内没有完成，则这次健康检查视为失败。--start-period=5s: Docker 会等待容器启动 5 秒钟 后，才开始执行第一次健康检查。在这 5 秒内的失败不会计入重试次数。--retries=3：如果连续 3 次 健康检查都失败，Docker 就会把容器状态设为 unhealthy。docker build -t flask-health .docker run -d -p 5000:5000 --name flask_app flask-health启动过程日志：
## 刚启动(health: starting)
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
193c5b0b3106 flask-health "python app.py" 10 seconds ago Up 8 seconds (health: starting) 0.0.0.0:5000->5000/tcp
## 一段时间后(healthy)
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES193c5b0b3106 flask-health "python app.py" 13 seconds ago Up 11 seconds (healthy) 0.0.0.0:5000->5000/tcp flask_app
docker-compose up -d命令时，为什么会出现depends_on依赖关系看似未生效的情况？depends_on 的真正作用: depends_on 只控制容器的启动顺序，不保证被依赖服务就绪可用。depends_on 作用验证：
FROM cr.loongnix.cn/library/debian:buster
RUN apt update && apt install curl -y
ENV TARGET_URL=http://flask_app:5000/health
CMD if curl --connect-timeout 3 --max-time 3 -fsS "$TARGET_URL"; then \
echo "Health check succeeded! Container will stay alive."; \
tail -f /dev/null; \
else \
echo "Health check FAILED. Exiting."; \
exit 1; \
fi
docker build -t health-probe .- docker-compose.yml
version: "2.10"
services:
flask_app:
image: flask-health
ports:
- "5000:5000"
probe:
image: health-probe
depends_on:
- flask_app
输出日志：
## `flask-health` 没有进入healthy `health-probe` 启动
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
a892e098fe9f health-probe "/bin/sh -c 'if curl…" 4 seconds ago Up Less than a second test-probe-1
9f86b9828cde flask-health "python app.py" 5 seconds ago Up 1 second (health: starting) 0.0.0.0:5000->5000/tcp test-flask_app-1
## probe 服务日志
curl: (7) Failed to connect to flask_app port 5000: Connection refusedHealth check FAILED. Exiting.
depends_on.condition: service_healthy ?depends_on.condition: service_healthy是 Docker Compose 的一个配置选项，用于指定一个服务（Service）只有在另一个服务健康状态正常时才会启动。它通常与 healthcheck配置一起使用。
核心作用:
具体修改
< - flask_app
---
> flask_app:
> condition: service_healthy
启动日志：
## 1. `flask-health` 进入healthy之前 `health-probe` 是不会启动的
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
7ba743998c8a flask-health "python app.py" 19 seconds ago Up 15 seconds (health: starting) 0.0.0.0:5000->5000/tcp
## 2. `flask-health` 进入healthy状态 `health-probe` 启动
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
ceb9071ac270 health-probe "/bin/sh -c 'if curl…" 49 seconds ago Up 25 seconds test-probe-1
7ba743998c8a flask-health "python app.py" 50 seconds ago Up 47 seconds (healthy) 0.0.0.0:5000->5000/tcp
## health-probe 日志
{"status":"ok"}Health check succeeded! Container will stay alive.
50c50,51
< - log
---
> log:
> condition: service_healthy
81c82,83
< - log
---
> log:
> condition: service_healthy
106c108,109
< - log
---
> log:
> condition: service_healthy
143,146c146,153
< - log
< - registry
< - redis
< - postgresql
---
> log:
> condition: service_healthy
> registry:
> condition: service_healthy
> redis:
> condition: service_healthy
> postgresql:
> condition: service_healthy
171c178,179
< - log
---
> log:
> condition: service_healthy
202c210,211
< - core
---
> core:
> condition: service_healthy
224c233,234
< - log
---
> log:
> condition: service_healthy
252,255c262,269
< - registry
< - core
< - portal
< - log
---
> registry:
> condition: service_healthy
> core:
> condition: service_healthy
> portal:
> condition: service_healthy
> log:> condition: service_healthy
修改之后，使用docker-compose up -d 命令可以稳动启动
1. depends_on仅确保容器按指定顺序启动，但不会验证被依赖服务的实际可用性。若部署时需要确保依赖服务已完全就绪（如数据库初始化完成），则需使用 depends_on.condition: service_healthy 对服务进行显式声明。