[nova] Can we drop the cells v1 docs now?
I have kind of lost where we are on dropping cells v1 code at this point, but it's probably too late in Stein. And technically nova-network won't start unless cells v1 is configured, and we've left the nova-network code in place while CERN is migrating their deployment to neutron*. CERN is running cells v2 since Queens and I think they have just removed this [1] to still run nova-network without cells v1.
There has been no work in Stein to remove nova-network [2] even though we still have a few API related things we can work on removing [3] but that is very low priority. To be clear, CERN only cares about the nova-network service, not the APIs which is why we started removing those in Rocky.
As for cells v1, if we're not going to drop it in Stein, can we at least make incremental progress and drop the cells v1 related docs to further signal the eventual demise and to avoid confusion in the docs about what cells is (v1 vs v2) for newcomers? People can still get the cells v1 in-tree docs on the stable branches (which are being published [4]).
[1] https://github.com/openstack/nova/blob/bff3fd1cd/nova/cmd/network.py#L43 [2] https://blueprints.launchpad.net/nova/+spec/remove-nova-network-stein [3] https://etherpad.openstack.org/p/nova-network-removal-rocky [4] https://docs.openstack.org/nova/queens/user/cells.html#cells-v1
*I think they said there are parts of their deployment that will probably never move off of nova-network, and they will just maintain a fork for that part of the deployment.
On Mon, Feb 11, 2019 at 9:43 AM Matt Riedemann mriedemos@gmail.com wrote:
I have kind of lost where we are on dropping cells v1 code at this point, but it's probably too late in Stein. And technically nova-network won't start unless cells v1 is configured, and we've left the nova-network code in place while CERN is migrating their deployment to neutron*. CERN is running cells v2 since Queens and I think they have just removed this [1] to still run nova-network without cells v1.
There has been no work in Stein to remove nova-network [2] even though we still have a few API related things we can work on removing [3] but that is very low priority. To be clear, CERN only cares about the nova-network service, not the APIs which is why we started removing those in Rocky.
As for cells v1, if we're not going to drop it in Stein, can we at least make incremental progress and drop the cells v1 related docs to further signal the eventual demise and to avoid confusion in the docs about what cells is (v1 vs v2) for newcomers? People can still get the cells v1 in-tree docs on the stable branches (which are being published [4]).
I think from an operators perspective, the documentation should at least be ripped out (and any nova-manage commands, assuming there's any). I guess there should be any tooling to allow you to get a cells v1 deployment (imho).
Cells V2 have been out for a while, extensively tested and work pretty well now.
[1] https://github.com/openstack/nova/blob/bff3fd1cd/nova/cmd/network.py#L43 [2] https://blueprints.launchpad.net/nova/+spec/remove-nova-network-stein [3] https://etherpad.openstack.org/p/nova-network-removal-rocky [4] https://docs.openstack.org/nova/queens/user/cells.html#cells-v1
*I think they said there are parts of their deployment that will probably never move off of nova-network, and they will just maintain a fork for that part of the deployment.
--
Thanks,
Matt
+1 to remove cellsV1 docs. This architecture should not be considered in new Nova deployments.
As Matt described we use cellsV2 since Queens but we are still using nova-network in a significant part of the infrastructure.
I was always assuming that cellsV1/nova-network code would be removed in Stein. I continue to support this plan! We will not maintain an internal fork but migrate everything to Neutron.
Belmiro CERN
On Mon, Feb 11, 2019 at 3:44 PM Matt Riedemann mriedemos@gmail.com wrote:
I have kind of lost where we are on dropping cells v1 code at this point, but it's probably too late in Stein. And technically nova-network won't start unless cells v1 is configured, and we've left the nova-network code in place while CERN is migrating their deployment to neutron*. CERN is running cells v2 since Queens and I think they have just removed this [1] to still run nova-network without cells v1.
There has been no work in Stein to remove nova-network [2] even though we still have a few API related things we can work on removing [3] but that is very low priority. To be clear, CERN only cares about the nova-network service, not the APIs which is why we started removing those in Rocky.
As for cells v1, if we're not going to drop it in Stein, can we at least make incremental progress and drop the cells v1 related docs to further signal the eventual demise and to avoid confusion in the docs about what cells is (v1 vs v2) for newcomers? People can still get the cells v1 in-tree docs on the stable branches (which are being published [4]).
[1] https://github.com/openstack/nova/blob/bff3fd1cd/nova/cmd/network.py#L43 [2] https://blueprints.launchpad.net/nova/+spec/remove-nova-network-stein [3] https://etherpad.openstack.org/p/nova-network-removal-rocky [4] https://docs.openstack.org/nova/queens/user/cells.html#cells-v1
*I think they said there are parts of their deployment that will probably never move off of nova-network, and they will just maintain a fork for that part of the deployment.
--
Thanks,
Matt
participants (3)
-
Belmiro Moreira
-
Matt Riedemann
-
Mohammed Naser