Service[nginx] seems to have an exec that fails due to being an empty string

[CpuID]

Not sure how this is actually happening, but reporting here for further input. Puppet 3.4.2.

...
Info: Concat[/etc/nginx/sites-available/api_internal_http.conf]: Scheduling refresh of Class[Nginx::Service]
Debug: Concat[/etc/nginx/sites-available/api_internal_http.conf]: The container Nginx::Resource::Vhost[api_internal_http] will propagate my refresh event
Info: Class[Nginx::Service]: Scheduling refresh of Service[nginx]
Debug: /Stage[main]/xxxxx::Nginx/Nginx::Resource::Vhost[api_internal_http]/File[api_internal_http.conf symlink]/mode: Not managing symlink mode
Notice: /Stage[main]/xxxxx::Nginx/Nginx::Resource::Vhost[api_internal_http]/File[api_internal_http.conf symlink]/ensure: created
Info: /Stage[main]/xxxxx::Nginx/Nginx::Resource::Vhost[api_internal_http]/File[api_internal_http.conf symlink]: Scheduling refresh of Service[nginx]
Debug: /Stage[main]/xxxxx::Nginx/Nginx::Resource::Vhost[api_internal_http]/File[api_internal_http.conf symlink]: The container Nginx::Resource::Vhost[api_internal_http] will propagate my refresh event
Debug: Nginx::Resource::Vhost[api_internal_http]: The container Class[Kix_vc_components::Nginx] will propagate my refresh event
Debug: Service[nginx](provider=upstart): Could not find nginx.conf in /etc/init
Debug: Service[nginx](provider=upstart): Could not find nginx.conf in /etc/init.d
Debug: Service[nginx](provider=upstart): Could not find nginx in /etc/init
Debug: Executing '/etc/init.d/nginx status'
Debug: Executing '/etc/init.d/nginx status'
Debug: Executing ''
Error: /Stage[main]/Nginx::Service/Service[nginx]: Failed to call refresh: Could not restart Service[nginx]: Execution of '' returned 1: 
Error: /Stage[main]/Nginx::Service/Service[nginx]: Could not restart Service[nginx]: Execution of '' returned 1: 
Debug: Class[Kix_vc_components::Nginx]: The container Stage[main] will propagate my refresh event
Debug: Finishing transaction 36376560
Debug: Storing state
Debug: Stored state in 0.28 seconds

[CpuID]

Also, this is Ubuntu 12.04 Precise amd64.

Results of /etc/init.d/nginx status:

[xxxxx] root@blah:~# /etc/init.d/nginx status
 * nginx is running
[xxxxx] root@blah:~# echo $?
0

[jfryman]

Woah. That's wild. I've never seen such a thing. Let me ask a few things to put this in context:

• Did this just start happening, a new install (of puppet, of nginx, of the machine itself?).
• Are the manifests in that trigger this run successfully in other places?

I'd love to get more data, but I really don't even quite know where to start attacking this since it's just... well... odd. My gut says this is a weird puppet parsing error we just need to suss out. I'll also setup an environment on my end tomorrow morning to see if I can reproduce this, but any info you can share in the meantime would be super helpful.

[CpuID]

Yea... first time for me also.

These machines are AWS instances, that I can spin up and tear down with 10~mins notice (100% ephemeral). I believe the first puppet catalog run actually succeeds without any issues, and only when a refresh is scheduled due to a config file change does this issue occur.

I am currently running my fork from https://github.com/CpuID/puppet-nginx which is about 10 commits behind your master branch. I reviewed the commits between our branches and could not see anything close to relevant here... but I am happy to test your master branch to verify if it's deemed relevant.

I can successfully restart nginx manually, without any issues after puppet completes.

This is currently the only failure throughout my catalog run, all other resources execute successfully.

[CpuID]

I think I may have a source of what is causing this, I have configtest_enable turned on.

service.pp:

 25   service { 'nginx':
 26     ensure     => running,
 27     enable     => true,
 28     hasstatus  => true,
 29     hasrestart => true,
 30   } 
 31   if $configtest_enable == true {
 32     Service['nginx'] {
 33       restart => $service_restart,
 34     }
 35   }

Is it possible that if lines 32/33/34 are executed, the actual command called is empty?

I can confirm that any config file changes that trigger a refresh cause this exit code 1 for me right now, reproducible.

[CpuID]

OK - I take this back entirely, I was setting service_restart to an empty string :| User error, move along... haha.