Re: rails app deployment

[Ezra Zygmuntowicz <ez@engineyard.com>]

	I assign a node["instance_role']  attribute for each of my nodes and  
only the 'app_master' role runs the migrations, the other app servers  
are 'app_slaves'.

	So basically you just check what role you are and set the migrate  
attribute of the deploy resource. to true or false depending on  
whether or not you want it to run migrations. Easy to script around.

        There are also callbacks that you put in the app being deployed:

APP_ROOT/deploy/{before_migrate.rb,before_restart.rb,before_symlink.rb}

	These ruby files will be called at the appropriate time in the deploy  
process and will be passed in the RAILS_ENV(or RACK_ENV,MERB_ENV) and  
the role of the instance as ARGV[0] and ARGV[1].

	You can write any kind of ruby code in these files to accomplish  
whatever you need your hooks to accomplish.


	I'd love to see a tarball version of this resource, I already have  
git done and I have SVN almost done, tarball would be the final one to  
do and I don't forsee wanting to support any further SCM's as I want  
to keep this fairly simple.

Cheers-
-Ezra


On Apr 17, 2009, at 10:17 AM, Ian Kallen wrote:

> Oh, sweet, that's pretty much what I was imagining. Do you have  
> plans to support "push" deployment (roll an app tarball, dist out to  
> servers and shimmy the symlinks) ? Also (rails newbness alert) - are  
> there race conditions with db migrates when you have a cluster of  
> machines updating in parallel that risk fubaring the schema? I can  
> imagine without a global lock asserted by the first cluster member  
> to start a migration, you could get cluster members stepping on each  
> other trying to execute migrations in parallel.
> thanks!
> -Ian
>
> Ezra Zygmuntowicz wrote:
> >    I've written chef deploy resources that handle this. The only  
> > thing not implemented yet is rollback but that will be done soon.
> >
> > http://github.com/ezmobius/chef-deploy/tree
> >
> >
> > Cheers-
> > -Ezra
> >
> > On Apr 17, 2009, at 9:18 AM, Ian Kallen wrote:
> >
> > > Sorry if this is an FAQ I've overlooked. I'm trying to get my head  
> > > around best practices wrt to rails deployment. To manage  
> > > deployment/rollback of a rails app to a set of apache/passenger  
> > > instances, would you create a recipe that invokes capistrano  
> > > tasks, create cookbook that gems up the app, distributes and  
> > > installs (how would rollbacks be handled?). I see the rails  
> > > cookbook on github but it's not clear what the over all workflow  
> > > is - if the rails code is housed in git, how do you specify  
> > > "deploy this new revision" or "rollback to that revision"? How are  
> > > db migrations handled?
> > > thanks!
> > > -Ian
> > >
> > > --
> > > Ian Kallen
> > > blog: http://www.arachna.com/roller/spidaman
> > > tweetz: http://twitter.com/spidaman
> > > vox: 415.505.5208
>
>
> --
> Ian Kallen
> blog: http://www.arachna.com/roller/spidaman
> tweetz: http://twitter.com/spidaman
> vox: 415.505.5208

Ezra Zygmuntowicz
ez@engineyard.com