VMware Virtualization for Mac Beta released!

I've been waiting for the ability to use the same VMware guest images on my linux servers as my MacBook Pro. This means I can replicate my production servers on my laptop, removing the need to have direct access to any other computer. Just in time for India!



I was able to point it at copy of an ubuntu image currently running in production and it booted without a problem. Nice! Not all the options for creating a new image are available in this release.

Download it here.

Setting environment variables for mongrel

We were going to use Verisign's PayFlowPro merchant facilities and implemented a solution based around a ruby library called pfpro4r. A problem came up though. We needed to have a shell environment variable set in order for this to work. This was fine when I started mongrel manually but when started via Capistrano the environment variable wasn't being set by the mongrel start script.

You can't seem to set environment variables in the shell Capistrano uses so I simply added a couple of tasks to deploy.rb. These override the tasks that come with mongrel_cluster and set the required variables.

I think it's great that you can add or override tasks on a project by project basis.
I guess someone distributing their app could include a library like deprec and include it from their deploy.rb to make deployment that much easier.

desc <<-DESC Start Mongrel processes on the app server. This uses the :use_sudo variable to determine whether to use sudo or not. By default, :use_sudo is set to true. DESC task :start_mongrel_cluster , :roles => :app do
 set_mongrel_conf
 send(run_method, "sh -c '
 LD_LIBRARY_PATH=/var/www/apps/example_app/current/vendor/libs/pfpro4r/verisign/payflowpro/linux/lib;
 export LD_LIBRARY_PATH;
 PFPRO_CERT_PATH=/var/www/apps/example_app/current/vendor/libs/pfpro4r/verisign/payflowpro/linux/certs;
 export PFPRO_CERT_PATH;
 mongrel_rails cluster::start -C #{mongrel_conf}
 '
 ")
end

desc <<-DESC
Restart the Mongrel processes on the app server by starting and stopping the cluster. This uses the :use_sudo variable to determine whether to use sudo or not. By default, :use_sudo is set to true.
DESC
task :restart_mongrel_cluster , :roles => :app do
  set_mongrel_conf

  send(run_method, "sh -c '
  LD_LIBRARY_PATH=/var/www/apps/example_app/current/vendor/libs/pfpro4r/verisign/payflowpro/linux/lib;
  export LD_LIBRARY_PATH;
  PFPRO_CERT_PATH=/var/www/apps/example_app/current/vendor/libs/pfpro4r/verisign/payflowpro/linux/certs;
  export PFPRO_CERT_PATH;
  mongrel_rails cluster::restart -C #{mongrel_conf}
  '
  ")
end

ssh compression can slow you down

[I've reposted this post on ssh compression to my new blog as I keep using it myself]

When copying some large files between hosts at our rack I noticed some pretty poor transfer speeds. Having recently forked out $70 for a rack mount gigabit switch I wondered what was slowing things.

It seems ssh was trying to help out by compressing everything however compressing the data took more than twice as long as transferring the uncompressed data. I proved this by disabling compression at the commandline and seeing the transfer speed more than triple.

mbailey@modu1:~/vm$ scp -r r03 modu4:vm/
Ubuntu 64-bit-f001.vmdk                         7%  147MB   8.1MB/s   03:54 ETA

mbailey@modu1:~/vm$ scp -r -o 'Compression no'  r03 modu4:vm/
Ubuntu 64-bit-f001.vmdk                       100% 2048MB  28.1MB/s   01:13 ETA

So how can we put this into practice everywhere?

Setting 'Compression no' in /etc/ssh/ssh_config will do the trick for me. There's no need for my vmware hosts to be compressing files I copy between them. I do want compression when copying files over shared/slower links but I can achieve that by initiating connections that benefit from compression on boxes that are configured to use it.

If I want to enable compression I can always use the '-C' switch to enable compression.

deprec - install rails stack and deploy in minutes

Update 2007-06-04: Visit www.deprec.org


Until now, installing a full Rails stack on Ubuntu linux and deploying your app took more than a few minutes. The first time I tried it took hours. Even copying and pasting from my notes took forever.

Now with deprec (deployment recipes for Capistrano) you can install a rails stack and deploy your application with just 6 commands.

cd /path/to/railsapp
deprec --apply-to . --name projectname --domain www.projectname.com
# edit config/deploy.rb to put in details of your subversion repository
cap install_rails_stack
cap deprec_setup
cap deploy_with_migrations
cap restart_apache

That's all it takes to get your Rails app running on a default Ubuntu 6.06 server install. Install the gem with:

sudo gem install deprec -y    # installs what you need
deprec_dotfiles               # patches capistrano + creates your ~/.caprc
cap show_tasks                # now you have deprec tasks included

I'll be putting more details up soon on deprec.rubyforge.org

Deprec was inspired and uses the brilliantly executed Capistrano. Thanks Jamis!

After starting on this project I found myself reading and utilizing a lot of
code by Bradley Taylor, in particular the RailsMachine and VMbuilder gems.
Neil Wilson wrote three capistrano plugins that are used in this project.
I'd like to say a huge thanks to these guys for helping make my work easier!