Question

I had my little rails-app (based on refinerycms) well up and running when I suddenly realized it had a small bug. Fixed it, tested ok, commit, push, deploy, the regular routine. However this time my deployment (which had worked countless times before) crashed on me during assets:precompile with

    * 2012-12-21 11:44:59 executing `deploy:assets:precompile'
    * executing "cd /home/somename/releases/20121221104442 && bundle exec rake RAILS_ENV=production RAILS_GROUPS=assets assets:precompile"
      servers: ["ssh.alwaysdata.com"]
      [ssh.alwaysdata.com] executing command
  *** [err :: ssh.alwaysdata.com] rake aborted!
  *** [err :: ssh.alwaysdata.com] Invalid argument - /nfs/http9/somename/app/ttvz/releases/20121221104442/tmp/cache/assets/D17/790/sprockets%2F159b50b9f479a8cb2519ea6a40f09f92
  *** [err :: ssh.alwaysdata.com] /home/somename/app/ttvz/shared/bundle/ruby/1.9.1/gems/activesupport-3.2.1/lib/active_support/core_ext/file/atomic.rb:38:in `chown'
  ...

and of course a whole bunch of a Stacktrace. I tried to find something on the web but failed miserably. The same thing also happened when I tried to precompile the assets of the current version of the app (which worked like a charm in development).

This seems to be different from the other precompile-troubles on StackOverflow.

Answer 1

Contrary to public belief this turned out not to be my error. I started looking up the source of atomic.rb and figured that the call to chown would actually end up doing exactly nothing. So I patched it away not without leaving a message in the alwaysdata user forum. With my (ugly) little patch deployments worked as before again.

A few hours later a friendly staff member told me that this lead him to suspect a recent kernel upgrade gone awry. He fixed it the same evening and everything is working again (without my patch).