Closed pjammer closed 8 years ago
if you look on my fork, which is way too messy to bring over right now, I had to add the code here, and add the symlink to attributes/default.rb
reason why is reference in i #25 but it works!
I'm afraid, I don't get it. The environmental variables are passed to unicorn (currently the only supported app-server), and they should be available out-of-the box in the ENV
object.
Why add another layer of abstraction?
When you have a bunch of ENV variables, like AWS_SECRET, or STRIPE... those are cool to keep in Opsworks in their console; but to get them into the app during deploy, you have to use (if you don't role your own) a gem like Figaro which takes an application.yml
file and allows you to use them in the app.
As an aside: I did notice whilst doing this that all ENV variables that you enter into the Opsworks console are also passed, along with RAILS_ENV=production
and it mucks up a stage for migration when you try to symlink_before_migrate
in a custom json stanza in the console.
Removing that custom json and not using a symlink_before_migrate
worked.
So I'm all ears if I did this thing wrong, and you have a better way man. I was in a pinch and this is what i had to do to get past it.
Can you explain me in layman's terms, with examples, what exactly do you try to achieve? Sorry, I still don't get it...
hey @ajgon and @pjammer,
I get what pjammer wants. Basically if you're booting up a process that's not unicorn, you'll want all the environment variables.
Right now you only get that directly booting unicorn. https://github.com/ajgon/opsworks_ruby/blob/master/templates/default/unicorn.conf.erb#L23 This will get run on deploy.
I think the advantages to having a yml are obvious. we dont need to insert the database configs into unicorn every time because its already available when the rails environment loads. If we want to spin up a non unicorn process like sidekiq it knows about the database too. Someone might want to ssh into a machine and boot up the rails console, they wouldn't need to enter in every variable into the command line.
@pjammer what you want is available already in open source projects. for example i've used this before https://github.com/joeyAghion/opsworks_custom_env it does exactly what you are describing. I haven't tested this for chef12 though.
making the case to have it included in this project, i don't know. I think its up to @ajgon but doing so would remove the need to inject environment vars into https://github.com/ajgon/opsworks_ruby/blob/master/templates/default/sidekiq.monitrc.erb#L10 and https://github.com/ajgon/opsworks_ruby/blob/master/templates/default/unicorn.conf.erb#L23
Ah, now I see it. I like the idea - added support for both figaro
and dotenv
- they are configurable via app['appserver']['application_yml']
and app['appserver']['dot_env']
stack options (both are turned off by default).
However I disagree, with a concept to remove those ENV for appserver/worker raw configuration files. The reason behind this is, that some people might not want/know/need to use them, and I want to make those scripts as less dependant on other gems as possible.
I can hardly wait to update my opsworks stack to this latest cookbook. I hate dependencies too, but for some reason what figaro
specially gives me I couldn't just make happen. i.e., the magic.
Did you happen to fix the fact that unicorn gets ALL the ENV
variables and i am getting this
> grep -c "WARNING" log/unicorn.stderr.log
493
example line is:
I, [2016-10-14T15:34:17.280463 #7590] INFO -- : reaped #<Process::Status: pid 7599 exit 0> worker=2
I, [2016-10-14T15:34:17.280587 #7590] INFO -- : reaped #<Process::Status: pid 7602 exit 0> worker=3
I, [2016-10-14T15:34:17.487023 #7590] INFO -- : reaped #<Process::Status: pid 7593 exit 0> worker=0
I, [2016-10-14T15:34:17.487132 #7590] INFO -- : reaped #<Process::Status: pid 7596 exit 0> worker=1
I, [2016-10-14T15:34:17.487218 #7590] INFO -- : master complete
I, [2016-10-14T15:34:18.344518 #6993] INFO -- : Refreshing Gem list
WARNING: Skipping key "DEVISE_SECRET_KEY". Already set in ENV..
I, [2016-10-14T15:34:24.194831 #6993] INFO -- : unlinking existing socket=/srv/www/wickedapp/shared/sockets/unicorn.sock
I, [2016-10-14T15:34:24.195029 #6993] INFO -- : listening on addr=/srv/www/wickedapp/shared/sockets/unicorn.sock fd=11
I, [2016-10-14T15:34:24.409746 #6998] INFO -- : worker=0 ready
I, [2016-10-14T15:34:24.549711 #7001] INFO -- : worker=1 ready
I, [2016-10-14T15:34:24.609199 #7004] INFO -- : worker=2 ready
I, [2016-10-14T15:34:24.627257 #6993] INFO -- : master process ready
I, [2016-10-14T15:34:24.707866 #7007] INFO -- : worker=3 ready
Basically 11 lines each time the reaper comes. or during a deploy... unclear as to why yet, but instead of just passing the UNICORN based envs, the whole kit'n caboodle comes over.
would be great to create an
application.yml
file with anyenvironment variables
you create in the Opsworks console. Then it'd be there if you use Figaro, or if you use dot.env maybe we put that format in there too. Right now, i'm struggling to figure out how to get those ENVs into this recipe so i can create a Factory for you. i tried even just cowboy coding it but ran up against time and talent.