[CentOS-devel] 'vagrant up' not working after yum update

Wed May 18 15:08:51 UTC 2016
Dominic Cleal <dominic at cleal.org>

On 18/05/16 15:52, Dharmit Shah wrote:
> Hello all,
> 
> I'm seeing this behaviour since I did `yum update` today on one of my 
> CentOS 7 boxes in that, `vagrant up` for libvirt provider throws the error:
> 
> "The provider 'libvirt' could not be found, but was requested to
> back the machine 'default'. Please use a provider that exists."
> 
> Details of the issue can be found on [1]. I observed the issue when I 
> was doing testing for Atomic Developer Bundle (ADB) [2].

Could you try this newer build of sclo-vagrant1-vagrant (the -5
release)? http://cbs.centos.org/koji/buildinfo?buildID=11113

It's been tagged for release already so I expect it'll appear on the
mirrors within a few days.

(see also https://www.redhat.com/archives/sclorg/2016-May/msg00014.html
where this was discussed and hopefully fixed)

> Besides that, earlier, one needed to install only two packages - 
> centos-release-scl and sclo-vagrant1 - to use Vagrant. But now it seems 
> to have changed and we need to install sclo-vagrant1-vagrant-libvirt as 
> well as few other packages/dependencies to make it work. I modified my 
> Ansible playbook [3] which reflects the set of packages I had to add.

This appears to have been changed in
http://cbs.centos.org/koji/buildinfo?buildID=10807, but I don't see why.
The changelog is pretty unhelpful, but comparing the specs shows that
%install_scl's been flipped to zero which isn't common in SCL meta RPMs.

> What I'm here to understand is that, am I doing something obviously 
> wrong here to make things work or, did I miss some notification on this 
> list that had details about the changes? Or is this some kind of 
> unexpected behaviour?

Unexpected to me, I think the RPMs should have been put to the testing
repo before being released.

-- 
Dominic Cleal
dominic at cleal.org