forward / capify-cloud Goto Github PK

We're using capify-cloud and it's working well. We're also using multistage and we've found a way to get it working with capify-cloud:

Our stage files look like:

@capify_cloud = CapifyCloud.new(fetch(:cloud_config, 'config/cloud.production.yml'))
cloud_roles :api

This allows us to load different configs for each stage where we're then overloading the project tag with an appended stage identifier:

:cloud_providers: ['AWS']
:project_tag: "projectname-staging"

:AWS:
  :aws_access_key_id: "ACCESS_KEY"
  :aws_secret_access_key: "SECRET_KEY"
  :regions: ["us-west-1", "us-east-1"]

This works, but it would be a lot slicker if we could tag stages rather than overload the project name.

To take this one step further, it would be good to be able to not have to duplicate config, as it's only the stage that changes. Specifying this directly in the stage file rather than creating new config files would be ideal.

I have the following ec2 setup:

server-1 Tag: Roles => "app"
server-2 Tag: Roles => "solr"

And have put the following in my config/deploy.rb:

cloud_roles :app, :solr

But when I execute the following command, capistrano only connects to the solr instance, not to both:

cap app solr shell
  * 16:10:56 == Currently executing `app'
  * 16:10:56 == Currently executing `solr'
  * 16:10:56 == Currently executing `shell'
cap> date
[establishing connection(s) to <server2-ip>]

Similarly, when I reverse the order (cap solr app shell), it connects to server-1, so it appears that the roles are overwriting each other rather than being additive.

As a workaround, I can use the following config:

cloud_roles :name => :app, :options=>{ :default => true }
cloud_roles :name => :solr, :options=>{ :default => true }

And then simply calling cap shell works as desired, but the documentation describing this configuration has a warning about this not working with load balancers, so I'm not sure if this is the correct approach.