Allow SSH synchronization between Docker containers

[nbendafi-yseop]

SUMMARY

When using Molecule, use of synchonize module works as expected when syncing data from controller to inventory_hostname (driver: Docker).

But if fails to sync data over SSH between two controlled containers during a play on one of them when the other is used as delegate_to

Note:

The two controlled host are prepared during Molecule prepare phase to communicate with each other (SSH keys shared and authorized_key updated accordingly)

ISSUE TYPE

• Bugfix Pull Request

COMPONENT NAME

lib/ansible/plugins/action/synchronize.py

ADDITIONAL INFORMATION

molecule --debug converge

---
- name: Converge
  hosts: server
  tasks:
    - name: Send data from backup server to destination server
      become: true
      synchronize:
        src: "/backup/{{ src_server }}/hourly.0/home/{{ user }}/{{ item }}"
        dest: "/home/{{ user }}/volumes//{{ item }}/"
      with_items:
        - conf
        - data
      delegate_to: "{{ backup_server }}"

failed: [server -> backup-server] (item=conf) => {
    "ansible_loop_var": "item",
    "changed": false,
    "cmd": "/bin/rsync --delay-updates -F --compress --archive --blocking-io --rsh=/usr/bin/docker exec -u root -i --out-format=<<CHANGED>>%i %n%L /backup/org-server/hourly.0/home/user/ server:/home/user",
    "invocation": {
        "module_args": {
            "_local_rsync_password": null,
            "_local_rsync_path": "rsync",
            "_substitute_controller": false,
            "archive": true,
            "checksum": false,
            "compress": true,
            "copy_links": false,
            "delete": false,
            "dest": "server:/home/user/volumes/conf/",
            "dest_port": null,
            "dirs": false,
            "existing_only": false,
            "group": null,
            "link_dest": null,
            "links": null,
            "mode": "push",
            "owner": null,
            "partial": false,
            "perms": null,
            "private_key": null,
            "recursive": null,
            "rsync_opts": [
                "--blocking-io",
                "--rsh=/usr/bin/docker exec -u root -i"
            ],
            "rsync_path": null,
            "rsync_timeout": 0,
            "set_remote_user": true,
            "src": "/backup/org-server/hourly.0/home/user/conf",
            "ssh_args": null,
            "times": null,
            "verify_host": false
        }
    },
    "item": "conf",
    "msg": "rsync: Failed to exec /usr/bin/docker: No such file or directory (2)\nrsync error: error in IPC code (code 14) at pipe.c(85) [sender=3.1.2]\nrsync: connection unexpectedly closed (0 bytes received so far) [sender]\nrsync error: error in IPC c
ode (code 14) at io.c(226) [sender=3.1.2]\n",
    "rc": 14
}

[nbendafi-yseop]

Looks like a test that has nothing to do with changed code has randomly failed, due to netwwork timeout

ipify_facts-yfyvNp / /root/ansible/test/results/.tmp/integration/ipify_facts-ocRFjZ-ÅÑŚÌβŁÈ/test/integration/targets/ipify_facts/tasks/main.yml:27 / [testhost] testhost: ipify_facts : get information about current IP using ipify facts validate_certs={{ validate_certs }}, timeout=30

{
"changed": false, 
"msg": "No valid or no response from url https://api.ipify.org/ within 30 seconds (timeout)"
}

(cf. Shippable #153281 - 78)

[samdoran]

Related to #62107 (comment).

[samdoran]

We will most likely not be adding further special casing to this plugin.

[nbendafi-yseop]

Thanks for the answer. But unlike #62107, which is clearly a new feature, this seems to be a bug fix. Having connection transport as Docker between controller and controlled node A and B shouldn't involve that connection transport between node A and node B must be Docker as well.
So falling back to use command module would be rather sad.

[nbendafi-yseop]

@samdoran Should I consider this issue as dead end ?

[samdoran]

rebuild_merge