I'm deploying a rails app to a VPS with capistrano, bundler and rvm.
Here is part of my deploy.rb
$:.unshift(File.expand_path('./lib', ENV['rvm_path']))
require "rvm/capistrano"
require "bundler/capistrano" # Load RVM's capistrano plugin.
set :rvm_type, :system
set :rvm_ruby_string, '1.9.2#gemset_name'
my .rvmrc
rvm --create use 1.9.2#gemset_name
When I logged into the server I noticed that the rvm gemset was created, however all the gems were installed in /shared/bundle/ruby/1.9.1/gems --not in the gemset(/usr/local/rvm/gemset)
I use RVM in development and I think it's great but when is time to deploy to production. what are the best practices? Is there a way to tell bundler to install the gems in the gemset?
Do I even need a gemset at all? (seems like bundler is already isolating gemsets), Am I missing something?
Please help me understand!
Thank you very much
I use RVM in development and production as well. However, while I use gemsets in development to separate gems between my rails projects, I only use RVM to install rubies on my production VPS and let Bundler handle the versions of my gems.
Using the bundler integration via require "bundler/capistrano" automatically sets some things up for bundler. You can see the code behind this in Bundlers Github page. The basic settings are so that bundle executes this command:
bundle install --gemfile Gemfile --path shared/bundle --deployment --quiet --without development test
As you can see, the --deployment and --path flags are given, which tells Bundler to bundle your gems with your application in the shared/bundle directory, and only use the versions specified in your Gemfile.lock (i.e. the versions that are working in development). Since the bundle directory is shared between deployments, I find it comparable to using RVM gemsets for each application, just easier.
If you still want to put gems in separate gemsets I'd recommend this tutorial from Darcy Laycock for further reading.
Related
I need to deploy a Rails 3/Phusion Passenger/Apache2 app to an Ubuntu 12.04 server without installing ANY gems on the server.
I can install any Debian packages.
How can I include all needed gems (including Rails, a couple of gems installed from git projects on github, and a bunch of regular gems installed from rubygems.org) inside my project so that, after deployment, Passenger will find Rails and all the gems and be happy?
(I know many people will probably be curious WHY I would want to do this. I can't talk about that, unfortunately. Sorry.)
Here are some things that have not worked. One of them might work if revised a little, I don't know:
Failed method 1:
On Ubuntu 12.04 dev machine. Ruby 1.9.3. Bundler 1.3.5. bundle package --all. Git commit.
On Ubuntu 12.04 server machine (gem installation from Internet not possible). Deploy project files and install packages (apache2, ruby1.9.3, ruby-bundler, rubygems, libapache2-mod-passenger which brings ruby1.8 with it). bundle -v gives Bundler version 1.0.15 (from ruby-bundler package). Make sure .bundle/config file from dev machine is also deployed. bundle install --local tries to download git gems from github, so fail.
Try using bundle package --all, which should save your gems into ./vendor/cache. You could also unpack the gems on your local environment with gem unpack --target vendor to save them in the vendor directory. Then you can give a location in the Gemfile so that your application can find the gems. Can you just not use the network to install gems, or are you really not allowed to install gems in any way?
Sources:
1,
2
Try using the pkgr tool, which can convert your Rails app and all dependencies into a single installable Debian package.
Download the bundler 1.3.5 gem from http://rubygems.org/downloads/bundler-1.3.5.gem and put it in vendor/
In an Ubuntu 12.04 vm, do bundle package --all. Git commit and push.
Clone on the Ubuntu server.
apt-get install each of apache2, ruby1.9.3, etc. (not rubygems or ruby-bundler. Both provide an old version of Rubygems and Bundler. A newer version of rubygems is included in the ruby1.9.3 package.)
Follow ALL of the instructions in section 2.3 at http://www.modrails.com/documentation/Users%20guide%20Apache.html#install_on_debian_ubuntu
Make sure you have the right Ruby and Rubygems versions set to the ruby and gem commands:
update-alternatives --install /usr/bin/ruby ruby /usr/bin/ruby1.9.3 200
update-alternatives --install /usr/bin/gem gem /usr/bin/gem1.9.3 200
(Phusion Passenger doesn't care about this, but it's probably a good idea for the gems we are about to install and for debugging stuff with the Rails console or whatever in the future.)
gem install --local vendor/bundler-1.3.5.gem
bundle install --local
Set up all your other Apache config...
service apache2 restart
I want to customize my development environment with a few extra gems. I am using bundler with rails 3.0.x. I have those extra gems in my local system and i dont want to add them into my Gemfile. How can i by pass bundler and require these gems in my console which was opened using bundle exec ?
I found one way to do it, though it's a little hacky.
$ gem install gem_name
$ gem which gem_name
# outputs <full path to gem>/lib/gem_name.rb
$ rails console
> $LOAD_PATH << "<full path to gem>/lib"
> require 'gem_name'
I'm doing this to play around with different Ruby performance gems, although it might just be easier to add it to your Gemfile, bundle, and then revert before you push up your changes.
I'm running a Rails 3.0.3 app using RVM (Ruby 1.8.7). When I run bundle install all the gems install and then I get this message Your bundle is complete! It was installed into ./rails. But the app isn't recognizing/see those gems. Not sure what I've done wrong that it's doing this. Let me know if you need any other information.
I bet you accidentally ran bundle install rails at one point, which configured your bundler to install your gems into the ./rails directory.
To reset that, remove the ./.bundle directory (or remove the BUNDLE_PATH line from ./.bundle/config). Then bundle install will put them in the right spot.
I'm using Ubuntu 11.10 and the terminal to install and run Rails. Here
is the process I've taken so far to setup Rails:
download and install Ruby 1.9.2 and Rails 3.1.0
-- I did this using sudo apt-get ruby1.9.1 and sudo gem install
rails
I made a new rails app using rails new path/to/app
I went to the new app directory and tried running rails server and got an
error about not having a JS ruby environment
I had to get a JS runtime environment for execjs so I downloaded 'therubyracer'
as well as 'therubyrhino' and added them to my gem file, one at a time like this:
gem
'therubyracer' then ran bundle install
After everything was successful with the install, I ran rails server again
-- with both runtime environments I have had similar errors:
Could not find libv8-3.3.10.4 in any of the sources
Run `bundle install` to install missing gems.
Could not find therubyrhino-1.73.0 in any of the sources
Run `bundle install` to install missing gems.
Bundle knows where these programs live, giving correct pathnames when I
enter bundle show libv8 or bundle show therubyrhino. They are both
in /usr/lib/ruby/gems/1.9.1/gems/_ where all the other gems are
located for the bundle install.
Does anyone know where this exception is coming up in the Rails source
code? Does anyone know how Rails is gathering the gems? Better yet,
anyone had this problem and know how to fix it?
Thanks so much,
Feav
Have you tried starting the server using bundle exec rails server?
This problem has been fixed in a newer version of Rails.
Using rvm I installed and am using Ruby 1.9.2p180.
I created a gemset called rails3tutorial and, using it, installed Rails 3.0.7, created "sample_app", then used bundle to install. Very nice.
Then I created a new gemset called rails310b1, installed Rails 3.1.0 Beta 1, created an app, and bundle install'd...but it turned out I was not actually switched to the rails310b1 gemset, and so I ended up installing to ruby-1.9.2p180's default gemset (is that #global?...). I then manually deleted all the gems that had been installed!
Then I switched to rails310b1, and bundle install'd, created an app, fired up the app -- golden.
Now I switch back to the rails3tutorial gemset, and cd to the app created using it, and I get errors when running rails commands like...
$ rails generate integration_test layout_links
/Users/paul/.rvm/gems/ruby-1.9.2-p180#rails3tutorial/bin/rails:19:in `load': no such file to load -- /Users/paul/.rvm/gems/ruby-1.9.2-p180#rails3tutorial/gems/rails-3.1.0.beta1/bin/rails (LoadError)
from /Users/paul/.rvm/gems/ruby-1.9.2-p180#rails3tutorial/bin/rails:19:in `<main>'
It's looking for /Users/paul/.rvm/gems/ruby-1.9.2-p180#rails3tutorial/gems/rails-3.1.0.beta1/bin/rails but why?! $PATH looks fine.
Any suggestions as to what is messed up and how to clean this up, or is it best if I wipe out all my gems/gemsets and start over?...
Thanks!
There is nice screencast with basics of RVM http://screencasts.org/episodes/how-to-use-rvm but it mostly comes to using rubies:
rvm install 1.9.3
cd ~/projects/my-app
touch Gemfile
rvm use --create --rvmrc 1.9.3#rails32
gem install bundler rails
bundle exec rails new .
bundle install
bundle exec rails generate integration_test layout_links
please note you need to prefix most of the commands with bundle exec there is my gem that should save you from this problem rubygems-bundler.
also make sure you use .rvmrc files for every project so when you switch dir in console the proper ruby environment is set, the --rvmrc switch used above will generate one for you.