[Openstack-operators] Using Openstack-ansible to deploy hosts on mutliple br-vxlan/br-vlan/br-provider/br-mgmt l3 networks
bpetit at b0rk.in
Tue Jan 31 15:53:04 UTC 2017
Thank you for this sample
I assume osa is not yet ready to handle n provider_networks blocs ?
Here is the openstack_user_config I use (based on the one you proposed):
Or is there something I missed ? I get this error when I run
openstack-ansible setup-infrastructure.yml --syntax-check :
Variable files: "-e @/etc/openstack_deploy/user_secrets.yml -e
ERROR! The file
marked as executable, but failed to execute correctly. If this is not
supposed to be an executable script, correct this with `chmod -x
(/opt/openstack-ansible/playbooks/inventory/dynamic_inventory.py) had an
execution error: No container or management network specified in user
Expected key=value host variable assignment, got: argparse
On 12/01/2017 15:28, Jesse Pretorius wrote:
> Hi Benoit,
> Some time ago I did sketched out a possible way of doing this using a single inventory for OSA: https://gist.github.com/odyssey4me/09d963776f8872f2562e477c5158a3e0
> This is an entirely untested mechanism and may expose some horrible assumptions we’ve made in the way we configure things, but I thought it’d be a way of implementing OSA in a pod-like deployment. All this does, in theory, is have the dynamic inventory implement the container IP’s in different CIDR’s.
> If someone puts together some sort of PoC for making this work and identifies the shortcomings in the current mechanisms then I’m quite sure there would be many interested parties in the community who’d love to pitch in to help resolve them. It would make for a great discussion at the PTG!
> IRC: odyssey4me
> On 11/15/16, 5:38 PM, "bpetit at b0rk.in" <bpetit at b0rk.in> wrote:
> I would like to deploy OpenStack (newton) on multiple management,
> overlay and vlan/uplink l3 networks, thanks to Openstack-Ansible.
> In other words, I would like to deploy Openstack on multiple compute and
> controller nodes that are on different subnets, assuming I do the
> routing configuration to permit to management network 1 to communicate
> with management network n, overlay network 1 to overlay network n,
> I attached a sample schema to illustrate my statement.
> To sum up, in this case a compute or controller node which is on
> management network 10.0.1.0/24 via br-mgmt can communicate with br-mgmt
> of another compute node on management network 10.1.1.0/24. The same is
> true from overlay network 10.0.2.0/24 to overlay network 10.1.2.0/24,
> Am I wrong if I understand that OSA is designed to handle only 1
> mangement subnet, 1 overlay subnet, etc... (storage subnet, ...) ?
> During my tests, specifying compute hosts in different networks, in
> openstack_user_config.yml, results in the apparently correct setup of
> the compute host "compute3" (Having previously setup correct static
> route on each hypervisors):
> ip: 10.0.1.4
> ip: 10.0.1.5
> ip: 10.1.1.4
> However, how to deal with it in cidr_networks: ? (I assume I can't now).
> During my tests, an example of what's lacking was that I had to manually
> configure static routes in the neutron_agents_container lxc provisionned
> by OSA.
> Is that kind of deployment possible ? If not possible, is this or could
> this be in the roadmap of the project ?
> Benoit Petit
More information about the OpenStack-operators