cap servername deploy isnt working when servername has multiple roles

[masha256]

The deploy:assets:symlink task fails when doing a single server deploy to a server that has multiple roles (including the role that matches assets_role).

At first my server had "Roles" set to "app,db" and got the below error. When I edited the "Roles" aws tag to only have 'app', the deploy worked. I have assets_role defined as :app. So it seems something is not registering the server under the 'app' role when it is listed as having multiple roles.

I made sure the following was also in my deploy.rb:

ec2_roles :name=>:app, :options => {:default=>true}
ec2_roles :name=>:db, :options => {:default=>true}

When running "cap server1 deploy", I get the following error:

*** [deploy:update_code] rolling back

• executing "rm -rf /home/ubuntu/apps/myapp/releases/20130918175119; true"
  servers: ["ecxxxxxxxx.compute-1.amazonaws.com"]
  [ecxxxxxxxx.compute-1.amazonaws.com] executing command
  command finished in 920ms
  `deploy:assets:symlink' is only run for servers matching {:roles=>#<Proc:0x007f9e648bdf08@/Users/machado/.rvm/gems/ruby-1.9.3-p392@leaddyno-app/gems/capistrano-2.15.5/lib/capistrano/recipes/deploy/assets.rb:42 (lambda)>, :except=>{:no_release=>true}}, but no servers matched

[Ryan Conway]

Just checking, is this issue occurring even with the latest version of the gem (1.5.1)? I'll take a look as soon as I can otherwise.

[masha256]

Sorry, yep I made sure I was using 1.5.1.

[Ryan Conway]

That's a pain then! Would you mind attaching some diagnostic information, such as the output from ec2:status (filter out IPs etc for your privacy) and also, whatever you can share from your deploy.rb? As soon as I have time, I'll look into reproducing this and see if I can find a solution :)

Thanks!

[Ryan Conway]

Hey there, just been checking into this some more, and wondered if this had anything to do with it? capistrano/capistrano#128

Although they claim to have fixed it, Capify-EC2 currently requires an older version of Capistrano I believe, so it might be worth trying to ordering suggestion in that issue?

[Ryan Conway]

Hey there, since it's been a while, I'm going to close this issue, but feel free to re-open if you're still having trouble!

[James Martin]

My experience with this issue was when a server belonged to multiple roles. Whenever I do;

cap production servername deploy

It'll only recognize the "last" "ec2_roles" attribute (in my case, it wasn't :app) and so I had similar errors to
@machadolab.

To debug I have a task called :dump

# For debugging purposes
task :dump do
  puts roles.inspect
 end

When using;

cap production deploy:dump - all roles were shown.
cap production servername deploy:dump - only the "last" role would appear.