无法使用docker compose启动依赖容器

Question

我有两个Python服务，即first_service和second_service。以下是 docker compose 文件：

version: '3'

services:
  first_service:
    build: ./first_service
    container_name: first_service
    depends_on:
      second_service:
        condition: service_healthy
    restart: on-failure
    links: 
        - second_service

  second_service:
    build: ./second_service
    container_name: second_service
    volumes:
      - /home/user/settings/condition/second_service:/app/second_service
      - /home/user/settings:/app/settings
    healthcheck:
      test: ["CMD-SHELL", "test -f /app/settings/archive.sh && grep -q 'saket' /app/settings/archive.sh || exit 1"]
      interval: 5s
      timeout: 10s
      start_period: 10s

当我运行

docker compose up -d

时，second_service 进入状态

Up 34 seconds (unhealthy)

，first_service 进入状态

Created

。当第二个服务在一段时间后变得健康时，即

Up 38 seconds (healthy)

，第一服务仍处于

Created

状态。

我尝试使用

depends_on

和

service_healthy

选项，但它没有完成动机。我希望一旦 secondary_service 在任何时间变得健康，first_service 就能启动并运行。它应该仅通过 docker compose 文件发生。

Answer 1

如果我重现你的示例，它看起来像

docker compose

退出：

$ docker compose up
[+] Running 3/2
 ✔ Network containers_default             Created                0.1s
 ✔ Container containers-second_service-1  Created                0.0s
 ✔ Container containers-first_service-1   Created                0.0s
Attaching to first_service-1, second_service-1
second_service-1  | darkhttpd/1.14, copyright (c) 2003-2022 Emil Mikulic.
dependency failed to start: container containers-second_service-1 is unhealthy

因为 compose 退出，它不再管理容器，所以即使

second_service

变得健康，也没有什么等待启动

first_service

。这显然是预期的行为，尽管我觉得这令人惊讶。

解决方案是增加您的

start_period

，以便

second_service

在

start_period

到期时保持健康。例如：

  second_service:
    build: ./second_service
    container_name: second_service
    volumes:
      - /home/user/settings/condition/second_service:/app/second_service
      - /home/user/settings:/app/settings
    healthcheck:
      test: ["CMD-SHELL", "test -f /app/settings/archive.sh && grep -q 'saket' /app/settings/archive.sh || exit 1"]
      interval: 5s
      timeout: 10s
      start_period: 30s