Nginx doesn't restart on boot with xenial

Question

Nginx doesn't restart on boot with xenial

ctaverne opened this issue 6 years ago · comments

Christophe TAVERNE commented 6 years ago

Hi

I don't understand. My ansible script has this issue since 2 days (with ubuntu xenial).
Nothing changed in my playbook.

Now it doesn't enable nginx service on startup.

I get:

$ sudo systemctl status nginx
● nginx.service - A high performance web server and a reverse proxy server
Loaded: loaded (/lib/systemd/system/nginx.service; disabled; vendor preset: enabled)
...

I need to make the following change to make it enabled

name: Ensure nginx is started and enabled to start at boot.
service:
name: nginx
state: started
enabled: yes
use: service
become: yes

I was inspired by: geerlingguy/ansible-role-varnish#65 and ansible/ansible#22303

I just updated ansible from 2.5.0 to 2.5.1 and role to 2.5.1 too.
Nginx repository updated recently stable version to 1.14 too..

But it remains the same with old versions.. :( I can't come back to a working situation

Jeff Geerling · Answer 1 · Fri Apr 20 2018 22:12:03 GMT+0800 (China Standard Time)

I haven't had an issue with this on any of my Xenial servers... do you know if you have any other strange configuration in place? E.g. are you overriding/removing systemd on these servers?

Christophe TAVERNE · Answer 2 · Fri Apr 20 2018 22:16:37 GMT+0800 (China Standard Time)

Hi,
First, thanks for your answer and for sharing this project.

Me too, I hadn't any issue until now

No, I'm using standard configuration. To be more specific, I'm using official AWS ubuntu/images/hvm-ssd/ubuntu-xenial-16.04-amd64-server-20180126 (ami-1b791862)

Jeff Geerling · Answer 3 · Wed Apr 25 2018 02:20:56 GMT+0800 (China Standard Time)

Very odd. I am using the same AMI in US-East-1, and have Nginx installed and working (even after reboot) without issue across a number of servers (right now, 8 instances with Ubuntu 16.04 and Nginx managed by the latest version of this role).

Jeff Geerling · Answer 4 · Wed Apr 25 2018 02:21:56 GMT+0800 (China Standard Time)

I'm inclined to think maybe it's an issue with Ansible itself? But if it were, I would be seeing it too. I'm running 2.5.1 currently and don't see any problem on a fresh test environment (nor in a test Docker container).

jportoles · Answer 5 · Thu Apr 26 2018 01:42:54 GMT+0800 (China Standard Time)

The official Xenial PPA for nginx 1.14 no longer auto enables the systemd job, I don't know if this is an issue with nginx itself or the PPA, but it's definitely not Ansible's fault.

Christophe TAVERNE · Answer 6 · Thu Apr 26 2018 07:35:58 GMT+0800 (China Standard Time)

@jportoles if nginx is no longer auto enabled in systemd, it doesn't explain why ansible can't manage this service anymore and can't enable it as previously?

Iain Houston · Answer 7 · Sun May 06 2018 20:10:31 GMT+0800 (China Standard Time)

Only after I do sudo systemctl enable nginx on the guest Ubuntu VM can I do vagrant reload to find nginx running on the rebooted VM.

geerlinggut/nginx has:

- name: Ensure nginx is started and enabled to start at boot.
  service: name=nginx state=started enabled=yes

and this should translate to systemctl enable nginx shouldn't it? (ansible 2.5.2)

So what do you think is going on?

On Guest :

Ubuntu latest 16.04 (geerlingguy/ubuntu1604)
nginx version: nginx/1.14.0

On host:

Vagrant 2.1.0
ansible 2.5.2

Jeff Geerling · Answer 8 · Mon May 07 2018 12:22:30 GMT+0800 (China Standard Time)

I wonder if this is an Ansible 2.5.x bug... ansible/ansible#39116

Christophe TAVERNE · Answer 9 · Mon May 07 2018 18:45:38 GMT+0800 (China Standard Time)

Hi,
I came back to 2.5.0 and 2.5.1 and got the same result. Maybe I lost something..

justin · Answer 10 · Wed Jun 06 2018 00:05:05 GMT+0800 (China Standard Time)

Just upgraded Ansible to 2.5.4 and it looks like this is fixed

Daniel Compton · Answer 11 · Mon Jun 11 2018 06:30:04 GMT+0800 (China Standard Time)

I have this issue too, though I got it in Ansible 2.5.3 but it still happened after rolling back to 2.4.3.0. I was previously able to run Ansible fine with 2.4.3.0 so I suspect there is something in a recent version of Nginx which has also changed here.

Daniel Compton · Answer 12 · Mon Jun 11 2018 11:14:35 GMT+0800 (China Standard Time)

Hmm, I ran this against 2.5.4 and still had the same issues with nginx not being enabled.

Jamin · Answer 13 · Sat Oct 27 2018 06:22:12 GMT+0800 (China Standard Time)

I'm on ansible 2.5.10 and the same issue has spread to a few of my other service roles besides nginx. (I implemented some handler workarounds)

Anyone know if this is still an issue beyond ansible 2.5?

On a side note, I think the actual problem is some idempotency testing in ansible returns false positives in both service and systemd modules when a package includes both sysv and systemd init files on a systemd enabled system. (Like Ubuntu 16.04)

Jamin · Answer 14 · Sat Oct 27 2018 07:06:51 GMT+0800 (China Standard Time)

It sorta looks like this was fixed in ansible 2.6 in this PR, so maybe if people still have this issue, that is the resolution. ansible/ansible#46245

Jeff Geerling · Answer 15 · Thu Nov 22 2018 10:34:09 GMT+0800 (China Standard Time)

Yeah; and in my quick testing with Drupal VM earlier today, it seemed to be working okay (I was testing a new base box which installed Nginx separately, then I rebooted and it was running correctly).

Closing this issue, feel free to reopen if you can reproduce the issue on newer versions of Ansible.