停止父容器时如何解除子容器的Docker网络关联？Ansible部署问题

Let’s break down why you’re hitting this issue first: Your pre_tasks try to stop/remove the main containers (and their attached network) before the role tasks can stop the sub-project containers that are still using that same network. Since Ansible runs pre_tasks first, then roles, sub-containers are still active when the main container’s docker_compose: state=absent tries to delete the network—hence the error.

Here are two reliable fixes you can implement:

Fix 1: Stop Sub-Containers Before Main Containers (Dynamic Pre-Task) #

Since you already have per-role stop logic, we can pull that into pre_tasks to run it before touching the main containers.

1. Extract stop tasks into a reusable file for each role
   For each project role (e.g., app-admin), create a tasks/stop_container.yml file with just the stop logic:

   ---
   - name: Stop existing {{ name }} docker containers
     docker_compose:
       project_src: "{{ dockerComposeFileDestination }}"
       state: absent
   

   Move the stop task from main.yml into this new file, then include it back in main.yml if you still need it there:

   # app-admin/tasks/main.yml
   ---
   - include_tasks: stop_container.yml
   # rest of your existing tasks...
   

2. Add a dynamic pre-task to stop sub-containers first
   In your main playbook’s pre_tasks, right before stopping the main containers, add this loop to stop all relevant sub-project containers:

   # Pre-tasks section in your main playbook
   - name: Stop active sub-project containers first
     include_tasks: "../roles/{{ item }}/tasks/stop_container.yml"
     loop: "{{ filteredProjectsList }}"
     when: item != 'all'  # Skip 'all' since we handle individual projects
     vars:
       name: "{{ item }}"
       dockerComposeFileDestination: "{{ lookup('vars', item + '_docker_path') }}"  # Match your role's var naming
   

   Note: Make sure role variables like dockerComposeFileDestination are accessible here—you might need to define them as global vars or pass them explicitly.

   Now the flow becomes:

   • Stop sub-project containers → removes their network endpoints
   • Stop main containers → network can be removed without errors

Fix 2: Mark the Network as External (Avoid Deleting It) #

If you don’t need to delete the network when stopping main containers, mark it as external in both main and sub-project docker-compose files. This tells Docker Compose not to manage the network’s lifecycle.

1. Create the network explicitly in Ansible
   Add this task to your pre_tasks (before starting any containers):

   - name: Ensure appnetwork exists
     docker_network:
       name: appnetwork
       state: present
   

2. Update main container's docker-compose.yml
   Set the network as external so Compose won’t try to delete it when stopping:

   networks:
     appnetwork:
       external: true
   

3. Keep sub-projects' docker-compose.yml as is
   You already set external: true here, so they’ll use the pre-created network without issues.

   With this setup, running docker_compose: state=absent on main containers only stops/removes the containers—not the network. Since the network isn’t being deleted, active endpoints from sub-containers won’t trigger an error.

Which Fix Should You Choose? #

• Use Fix 1 if you need full cleanup (including the network) every time you run the playbook.
• Use Fix 2 if you want to keep the network persistent (faster subsequent runs, as you don’t recreate the network each time).

内容的提问来源于stack exchange，提问作者Bogdan Dubyk