Unknown role ArgumentError

[Brian Takita]

I'm on a project where I'm trying to use a capistrano script for
different stages.
One stage has a utility role and does not have a database, web, app,
or memcache server. The other stages do.

I want to reuse some of the cap tasks that I use for a normal deploy,
however, when I try to deploy to the utility stage, I receive:
/usr/local/lib/ruby/gems/1.8/gems/capistrano-2.5.3/lib/capistrano/configuration/servers.rb:75:in
`role_list_from': unknown role `db_admin' (ArgumentError)

I would like to ignore the task, since the utility role is not
specified for that task.
Has anybody else run into this issue? What is the idiom to get around this?

Thank you,
Brian Takita

[Brian Takita]

To fix this, I implemented the following monkey-patches. However, I
don't want to be incompatible with the Capistrano idioms if I can
avoid it.

http://gist.github.com/57165

[Jamis Buck]

Sadly, Capistrano doesn't currently let you declare a role without
servers, but you can work around it like this:

roles[:db_admin]

Just that, on a line of it's own, will tell Capistrano that there is a
role named :db_admin. It won't have any servers in it (unless you later
do role(:db_admin,...)).

From there, you'll then need to make sure your tasks know not to try and
run anything if there are no matching servers (otherwise, cap complains
that no servers matched). Something like this:

task :something, :roles => :db_admin do
next if find_servers_for_task(current_task).empty?
# ...
end

And that should do it. In light of the hacks this requires, I'd be
willing to consider two patches:

1. allow role() to take only a role name, without servers, so that you
can "declare" an empty role.

2. add an option to task() so that it is skipped if there are no
matching servers, or which causes run() to be silent in that case.

- Jamis