Hi all,
I (probably inappropriately) got into a discussion about this one twitter earlier today. I wanted to try and get some understanding of the direction here because it does directly impact the viability of omnibus for us.
Previously in the original omnibus, ruby + omnibus was only required to create the project. A common workflow on our side was this:
- Clone project repo
- install vagrant-omnibus and vagrant-berkshelf plugins
- vagrant up
- collect and publish packages
Yes, I know that vagrant 1.5 broke that model because vagrant-berkshelf didn't work with it but we actually weren't upgrading BECAUSE of this. Understand, we're not a ruby shop. We use ruby very sparingly in our workflow. The only reason most of us have a ruby install is to use knife.
Now it seems that with Omnibus 3, the understanding is that a ruby environment will not only be required to create a project but to actually run the process to build the project. This apparently is due to moving to test-kitchen instead of vagrant straight.
This REALLY complicates having people involved internally in our omnibus builds to the point of probably having to drop using it.
I'm also curious as well because omnibus 3 brought in an unreleased dependency on Berkshelf 3.xbeta which also complicates things as there appears to be a move to have berkshelf be a requirement of everything.
So now I'm stuck in a weird spot. I need to have a ruby toolchain to run knife but now I need a different ruby toolchain to run omnibus (where I *DIDN'T* need one before) because every thing seems to be pulling in conflicting versions of Berkshelf. We're right back in the json mess again.
So I'd love to have someone help me understand if this is a definitive way forward for omnibus and if possible what the reasoning is behind making Berkshelf so core to everything.
Thanks!
John
Archive powered by MHonArc 2.6.16.