I have been looking at deploying a CentOS based tripleo proof of concept which could possibly become required to run in production. I understand deploying the undercloud and overcloud nodes (compute, controller and storage) are recommended to be installed on baremetal hosts. However, due to available resources and the possible need to have this virtualised I have been looking at deploying tripleo in a virtual environment (instack-virt-setup) following the below guide.
Is this kind of deployment agreed to be production capable?
As I understand it all the overcloud nodes reside on one physical host, however
is it possible to ensure resilience in this deployment - if this one physical node
goes down surely all the openstack environment is down?
Also in a baremetal environment scaling the overcloud would involve adding a new baremetal server updating ironic etc – how would this be possible in a virtual environment when during the deployment the node count has been set?
I would appreciate any feedback you can provide on these questions.