Ask Your Question

Revision history [back]

Hi there,

If you want to give H.A. only at controller level, you're OK, but, if you want to also give your instances (vm's) H.A., and also, live-migration, you need to either have a common shared space storage (NFS, GlusterFS) for /var/lib/nova/instances (the "instances" directory), or, use CEPH for nova and libvirt, so the actual storage storage for your instances (ephemeral, swap, etc.) is completely out of the compute nodes.

For the first option (/var/lib/nova/instances) you need a NFS or GlusterFS solution, so your "instances" directory will be stored in a NAS solution (again, NFS or GlusterFS): This means a separate equipment for the NAS.

For the second option (CEPH, which is by far one of the best options if not the very best for OpenStack), you need more hardware, but, you can escalate more !.

Also remember, if you plan to deploy Cloudfoundry into your OpenStack cloud, that you'll need at least 7 vm's for the cloudfoundry complete suit, and some 32 vcpu's and about 48-128 gigs of ram...bare minimum 48 gigs of ram.

If you go for the CEPH way, I have a recipe and complete LAB for Mitaka Series in my github-pages site:

http://tigerlinux.github.io

Good luck and success !