6

I have a systemd service that I deploy and want to be started by Ansible.

My systemd service unit file is this:

[Unit]
Description=Collector service
After=network.target mariadb.service
Requires=mariadb.service

[Service]
Type=simple
ExecStart=/opt/collector/app.py
WorkingDirectory=/opt/collector
Restart=on-abort
User=root

[Install]
WantedBy=multi-user.target

I am using Type=simple since this looks like the correct solution (also the preferred one in this Question).

I tried using Type=oneshot as well (as suggested by the initial user making this question as duplicate of this question) but the problem is that the /opt/collector/app.py script is a long running process:

while True:
    t = threading.Thread(...)
    t.start()
    t.join()
    time.sleep(15)

and with Type=oneshot, Ansible will block forever.

And my Ansible starting code is:

- name: start Collector service
  systemd:
    name: collector
    state: started
    enabled: yes

On the target system, systemctl will display:

[root@srv01 /]# systemctl
  UNIT                           LOAD   ACTIVE     SUB       DESCRIPTION
  dev-sda1.device                loaded activating tentative /dev/sda1
  -.mount                        loaded active     mounted   /
  dev-mqueue.mount               loaded active     mounted   POSIX Message Queue File System
  etc-hostname.mount             loaded active     mounted   /etc/hostname
  etc-hosts.mount                loaded active     mounted   /etc/hosts
  etc-resolv.conf.mount          loaded active     mounted   /etc/resolv.conf
  run-user-0.mount               loaded active     mounted   /run/user/0
  session-73.scope               loaded active     running   Session 73 of user root
  crond.service                  loaded active     running   Command Scheduler
  dbus.service                   loaded active     running   D-Bus System Message Bus
  haproxy.service                loaded active     running   HAProxy Load Balancer
<E2><97><8F> collector.service          loaded failed     failed   Collector service
....

The service fails because of the Python process exception (using un undefined variable).

But my Ansible playbook run does not fail:

TASK [inventory : start Collector service] *********************************
changed: [srv01]

I tried with both systemd and service Ansible modules and the behavior is the same.

How can I make Ansible:

  • fail when the systemd unit fails to start?
  • not block and systemd getting in active running status with a while True process?
Gabriel Petrovay
  • 18,575
  • 19
  • 87
  • 153
  • This question is not a duplicate of the mentioned question. See last edits. `oneshot` is not a solution due to the long running process that systemd needs to start. – Gabriel Petrovay Jul 29 '18 at 21:17
  • No one ever suggested `oneshot` was a solution for you, there are other options which you need to explore and choose. I closed the question because you asked for a reason why it did not return an error and the other answer explains that in a nice way. This is also not an Ansible problem, because `systemctl start` behaves exactly the same - so you can delete all that Ansible talk from your question. You are left with an open and too broad question "how to write a daemon in Python". – techraf Jul 30 '18 at 06:25
  • But the problem is that systemd alone behaves correctly. The difficulty is to have this running from Ansible (i.e. block forever or no showing the failures). If you have a solution how to solve this with Ansible, you can post an answer. Like that, thinking you know everything, you might block others from giving/receiving answers. Cheers! – Gabriel Petrovay Jul 30 '18 at 07:07
  • 1
    Also, thinking in a puritan way, I have no idea why you pointed out to another question that explicitly compares `simple` vs `oneshot` systemd options. For the "other options [...] to explore", I think you should have pointed me to: https://www.freedesktop.org/software/systemd/man/systemd.service.html (which by the way, I read but still not yet found a solution to make it run from Ansible) – Gabriel Petrovay Jul 30 '18 at 07:12
  • @GabrielPetrovay I pointed you to that answer because it contains a description of `Type=simple` and why it will never work for you with that setting.　・　So you claim `systemd` is behaving differently than Ansible and differently than how it was designed (read the other answer), right? And you call that "correctly"? Behaving contrary to documentation?　・　Attributing "*thinking you know everything*" attitude to someone who wanted to help is not only rude, but also counterproductive. – techraf Jul 30 '18 at 15:06

2 Answers2

2

I stumbled over this while I had the same problem with silently failing serives. I also found a bug report describing this issue and after some research I found a workaround:

- name: start Collector service
  systemd:
    name: collector
    state: started
    enabled: yes

- name: make sure Collector service is really running
  command: systemctl is-active collector

Note that for Type=simple services this will only fail if the service itself fails immediately after it was started.

harmathy
  • 21
  • 3
0

You can use failed_when example:

- name:  validating processes started correctly
  shell: pgrep toto| wc -l
  register: after_count
  failed_when: after_count.stdout_lines[0] == "1"

The failed_when will fail a task if the number of processes returned is not == 1

eduprado
  • 71
  • 1
  • 3