Handlers don't execute handlers in included

[galiev]

When I include a file in a handler, the tasks in the included file are not executed in ansible 2.0.
Playbook redis.yml:

---
- hosts: redis
  gather_facts: true
  vars:
    os_firewall_use_firewalld: false
    port_open: "{{ redis_port|default(6379) }}"
  tasks:
    - include: roles/firewall/tasks/firewall_open_port.yml
  handlers:
    - include: roles/firewall/handlers/main.yml

roles/firewall/handlers/main.yml

- name: save iptables
  shell: service iptables save

roles/firewall/tasks/firewall_open_port.yml

- include: iptables/iptables_open_port.yml
  when: not os_firewall_use_firewalld

For ansible 1.9.4:
roles/firewall/tasks/iptables/iptables_open_port.yml

- name: Open port
  shell: iptables -A INPUT -m conntrack --ctstate NEW -p tcp --dport "{{ port_open }}" -j ACCEPT
  notify:
    - save iptables

For ansible 2.0:
roles/firewall/tasks/iptables/iptables_open_port.yml

- name: Open port 
  iptables:
    chain: INPUT
    ctstate: NEW
    protocol: tcp
    destination_port: "{{ port_open }}"
    jump: ACCEPT
  notify:
    - save iptables

Result:

ansible --version
ansible 1.9.4
  configured module search path = None

PLAY [redis] ****************************************************************** 

GATHERING FACTS *************************************************************** 
ok: [redis-01-transport.shire.local]
ok: [redis-02-transport.shire.local]

TASK: [Open port] ********************************************* 
changed: [redis-01-transport.shire.local]
changed: [redis-02-transport.shire.local]

NOTIFIED: [save iptables] ***************************************************** 
changed: [redis-01-transport.shire.local]
changed: [redis-02-transport.shire.local]

PLAY RECAP ******************************************************************** 
redis-01-transport.shire.local : ok=3    changed=2    unreachable=0    failed=0   
redis-02-transport.shire.local : ok=3    changed=2    unreachable=0    failed=0   

 ansible --version
ansible 2.0.0
  config file = /home/rino/ansible/ansible.cfg
  configured module search path = Default w/o overrides

PLAY ***************************************************************************

TASK [setup] *******************************************************************
ok: [redis-01-transport.shire.local]
ok: [redis-02-transport.shire.local]

TASK [include] *****************************************************************
included: /home/rino/ansible/roles/firewall/tasks/firewall_open_port.yml for redis-01-transport.shire.local, redis-02-transport.shire.local

TASK [include] *****************************************************************
included: /home/rino/ansible/playbooks/../roles/firewall/tasks/iptables/iptables_open_port.yml for redis-02-transport.shire.local, redis-01-transport.shire.local

TASK [Open port] **********************************************************
changed: [redis-01-transport.shire.local]
changed: [redis-02-transport.shire.local]

PLAY RECAP *********************************************************************
redis-01-transport.shire.local : ok=4    changed=1    unreachable=0    failed=0   
redis-02-transport.shire.local : ok=4    changed=1    unreachable=0    failed=0

When I replace in playbook:

  handlers:
    - include: roles/firewall/handlers/main.yml

to

handlers:
  - name: save iptables
    shell: /usr/libexec/iptables/iptables.init save

it's work well. It's bug or my mistake?

[mindfreakthemon]

Can confirm the issue on ansible 2.1.0 on Centos 6.6.

[vagrant@platform ~]$ cat /etc/redhat-release
CentOS release 6.6 (Final)
[vagrant@platform ~]$ ansible --version
ansible 2.1.0
  config file = /etc/ansible/ansible.cfg
  configured module search path = Default w/o overrides

File roles/x/handlers/main.yml:

- include: mysql.yml

File roles/x/handlers/mysql.yml:

- name: import sql dump
  mysql_db: name={{mysql_db}} target={{db_dump_path}} state=import

File roles/x/tasks/main.yml:

- name: add mysql db
  mysql_db: name={{mysql_db}} encoding={{mysql_encoding}} collation={{mysql_collation}} state=present
  notify:
    - import sql dump

Not only the handler is not triggered, it does not throw any error if it does not exist.

[aaronk6]

We're also seeing this after upgrading from Ansible 1.9.4 to 2.0.0-0.7.rc2 on OS X 10.11.

[jimi-c]

Can you test devel or the stable-2.0 branch? This should be resolved.

[galiev]

Thank you for replay.
Tested version:

Name        : ansible
Version     : 2.1.0
Release     : 0.git201512180409.a391d6f.stable20.el7.centos

Playbook test.yml:

---
- hosts: lb
  gather_facts: false
  remote_user: root
  tasks:
    - include: roles/test/tasks/main.yml
  handlers:
    - include: roles/test/handlers/main.yml
    - name: restart chrony
      service:
        name: chronyd
        state: restarted

roles/test/tasks/main.yml

- name: Debug
  command: /bin/true
  notify:
    - sshd restart
    - restart chrony

roles/test/handlers/main.yml

- name: sshd restart
  service:
    name: sshd
    state: restarted

Result:

$ ansible-playbook test.yml -vv
Using /home/rino/ansible/ansible.cfg as config file
1 plays in test.yml

PLAY ***************************************************************************

TASK [include] *****************************************************************
included: /home/rino/ansible/roles/test/tasks/main.yml for lb-transport.shire.local

TASK [Debug] *******************************************************************
NOTIFIED HANDLER sshd restart
NOTIFIED HANDLER restart chrony
changed: [lb-transport.shire.local] => {"changed": true, "cmd": ["/bin/true"], "delta": "0:00:00.003088", "end": "2015-12-18 11:21:55.788825", "rc": 0, "start": "2015-12-18 11:21:55.785737", "stderr": "", "stdout": "", "stdout_lines": [], "warnings": []}

RUNNING HANDLER [restart chrony] ***********************************************
changed: [lb-transport.shire.local] => {"changed": true, "name": "chronyd", "state": "started"}

PLAY RECAP *********************************************************************
lb-transport.shire.local   : ok=3    changed=2    unreachable=0    failed=0    

it's didn't work out.

[mindfreakthemon]

Checked it on the devel branch - [still does not work if the handler is in another file which is included in the role's handlers/main.yml file]:

Name        : ansible
Arch        : noarch
Version     : 2.1.0
Release     : 0.git201512180409.a391d6f.devel.el6

File playbook.yml:

- name: apply common configuration to all nodes
  hosts: 127.0.0.1
  connection: local
  remote_user: vagrant
  roles:
    - common

File roles/common/handlers/main.yml:

- include: test.yml

File roles/common/handlers/test.yml:

- name: test handler
  debug: msg="Yay"

File roles/common/tasks/main.yml:

- name: test task
  command: /bin/true
  notify: test handler

Ansible output:

[root@platform ansible]# ansible-playbook /provisioning/playbook.yml -vvv
Using /etc/ansible/ansible.cfg as config file
 [WARNING]: provided hosts list is empty, only localhost is available

1 plays in /provisioning/playbook.yml

PLAY [apply common configuration to all nodes] *********************************

TASK [setup] *******************************************************************
ESTABLISH LOCAL CONNECTION FOR USER: root
127.0.0.1 EXEC ( umask 22 && mkdir -p "$( echo $HOME/.ansible/tmp/ansible-tmp-1450427408.28-142856526343450 )" && echo "$( echo $HOME/.ansible/tmp/ansible-tmp-1450427408.28-142856526343450 )" )
127.0.0.1 PUT /tmp/tmpZl9tcv TO /root/.ansible/tmp/ansible-tmp-1450427408.28-142856526343450/setup
127.0.0.1 EXEC LANG=en_US.UTF-8 LC_ALL=en_US.UTF-8 LC_MESSAGES=en_US.UTF-8 /usr/bin/python /root/.ansible/tmp/ansible-tmp-1450427408.28-142856526343450/setup; rm -rf "/root/.ansible/tmp/ansible-tmp-1450427408.28-142856526343450/" > /dev/null 2>&1
ok: [127.0.0.1]

TASK [common : test task] ******************************************************
task path: /provisioning/roles/common/tasks/main.yml:1
ESTABLISH LOCAL CONNECTION FOR USER: root
127.0.0.1 EXEC ( umask 22 && mkdir -p "$( echo $HOME/.ansible/tmp/ansible-tmp-1450427408.66-77735003354168 )" && echo "$( echo $HOME/.ansible/tmp/ansible-tmp-1450427408.66-77735003354168 )" )
127.0.0.1 PUT /tmp/tmpZl9tcv TO /root/.ansible/tmp/ansible-tmp-1450427408.66-77735003354168/command
127.0.0.1 EXEC LANG=en_US.UTF-8 LC_ALL=en_US.UTF-8 LC_MESSAGES=en_US.UTF-8 /usr/bin/python /root/.ansible/tmp/ansible-tmp-1450427408.66-77735003354168/command; rm -rf "/root/.ansible/tmp/ansible-tmp-1450427408.66-77735003354168/" > /dev/null 2>&1
NOTIFIED HANDLER test handler
changed: [127.0.0.1] => {"changed": true, "cmd": ["/bin/true"], "delta": "0:00:00.001711", "end": "2015-12-18 10:30:08.718368", "invocation": {"module_args": {"_raw_params": "/bin/true"}, "module_name": "command"}, "rc": 0, "start": "2015-12-18 10:30:08.716657", "stderr": "", "stdout": "", "stdout_lines": [], "warnings": []}

PLAY RECAP *********************************************************************
127.0.0.1                  : ok=2    changed=1    unreachable=0    failed=0

As you can see here, ansible indeed tries to notify the handler "test handler" but somehow it does not get included. If I move the handler from test.yml directly to main.yml it will get triggered successfully.

[booch]

I'm running into the same problem in v2.0.0-0.8.rc3. I was expecting to be able to break handlers into separate files, as I do with tasks.

If I put the handler in handlers/main.yml (within the role directory), it works fine. But if I try moving the handler to another file and including it from handlers/main.yml, the handler does not run. I get no error message about the include not working.

[jimi-c]

This is currently documented in the 2.0 known issues. This is a consequence of the fact that includes are now dynamic and not processed ahead of time, so Ansible doesn't know about tasks inside an include it hasn't processed yet.

[juodumas]

@jimi-c: could you please point to the document containing the known issues for 2.0?

Such usage (handlers/main.yml including other files) should be quite common with non-atomic roles, but maybe they are not, because I could not find many examples in the wild. Anyway, here are a few random roles that are now broken with 2.0:

• https://github.com/pjan/the-ansibles/tree/master/roles/mail/handlers
• https://github.com/pjan/the-ansibles/tree/master/roles/security/handlers
• https://github.com/jaywink/ansible-mailman-bundler/tree/master/handlers
• https://github.com/linagora/obm-deploy/tree/master/roles/obm-ui/handlers
• https://github.com/linagora/obm-deploy/tree/master/roles/opush/handlers

If handlers are not executed after all tasks complete, services will not be restarted and will stay running with old configuration. This can be a security risk and Ansible won't even warn about it.

[stemid]

Is there a workaround for this? I have opted to keep all my handlers centrally and include them in playbooks for use in playbooks, and in roles.

But now the only workaround seems to be that I re-define every handler inline in every playbook that needs them. Or is there another workaround besides downgrading back to 1.9?