On Aug 11, 2016 9:15 AM, "Philippe Lafoucrière" <philippe lafoucriere tech-angels com> wrote:
> Just for the records, we added a new node this week using the scaleup.yml playbook, and it went pretty well.
> We also upgraded from 1.2.0 to 1.2.1 along with a Centos Atomic upgrade, and it didn't went well :(
> All the images created by builders were "missing", and we had to rebuild everything in every project, leading to a long unavailability (hopefully during night).
This sounds like your registry was using ephemeral storage rather than being backed by a PV or object storage.
The docs provide some additional details for this if manually deploying the registry: https://docs.openshift.org/latest/install_config/install/docker_registry.html
If using openshift-ansible for deployment, the example inventory file provides some variables that allow for configuring an NFS volume, an OpenStack Cinder volume, or a s3 bucket: https://github.com/openshift/openshift-ansible/blob/master/inventory/byo/hosts.origin.example#L290
> So if you have a virtualization system above OS, you should definitely snapshot before each run...
> users mailing list
> users lists openshift redhat com