[announce][debian][wallaby] general availability of OpenStack Wallaby in Debian

Thomas Goirand zigo at debian.org
Thu Apr 15 19:55:21 UTC 2021


Hi!

It's my pleasure to announce the general availability of OpenStack
Wallaby in Debian.

I've just finished uploading everything to Debian Experimental today
(not in unstable, as Bullseye is frozen), and the Bullseye backports are
available the usual way, for example using extrepo (which is in the
official Debian backports):

apt-get install extrepo
extrepo enable openstack_wallaby
apt-get update
...

or directly setting-up the http://bullseye-wallaby.debian.net repository
the usual way.

Note that while Victoria is available on Buster and Bullseye (to ease
the transition), Wallaby is only available on Bullseye, which is due to
be released next month (if everything goes as planned, there's no
official release date decided yet if I understand well).

New in this release
===================

Masakari is now packaged. Masakari-dashboard is still waiting on the FTP
master NEW queue though.

A quick update on openstack-cluster-installer
=============================================

Wallaby can already be installed by OCI [1].

Note that OCI is now public-cloud ready, and can deploy specific nodes
for the billing (with cloudkitty):
- messaging (separate RabbitMQ cluster and Galera for Gnocchi)
- billmon / billosd (separate Ceph cluster for Gnocchi)

Our tests showed that this setup can scale to 10k+ VMs without any issue
(the separate Galera + RabbitMQ bus really helps) reporting 400+ metrics
per seconds. As always, OCI is smart enough so the additional nodes are
all optional (and not needed for smaller scales), and the cluster
reconfigures itself if you decide to add new node types in your deployment.

Thanks to new features added to puppet-openstack [2], the number of
uwsgi process adapts automatically to the number of cores available in
controller nodes.

Last, with OCI you may now enjoy a full BGP-to-the-host (over IPv6
un-numbered link local) networking setup. This also works with compute
nodes, as long as you decide to not use the DVR mode (if you do with to
use DVR, then you need L2 connectivity on the computes: that's a Neutron
"feature", unfortunately), or if you decide to use Neutron BGP routed
networking [3] (though this mode also still has some limitations at this
time, such as no support for virtual router external gateways).

In this setup, only the Network nodes need L2 connectivity to the
outside world. This also scales very nicely to *a lot* of nodes...
without any ARP spanning tree problems. We (at Infomaniak) now only use
this deployment mode in production due to its scalability.

Final words
===========

Please report any issue you may find, on OCI or on the Debian packages.

Cheers,

Thomas Goirand (zigo)

[1]
https://salsa.debian.org/openstack-team/debian/openstack-cluster-installer
[2]
https://review.opendev.org/q/topic:%22debian-uwsgi-support%22+(status:open%20OR%20status:merged)
[3]
https://docs.openstack.org/neutron/latest/admin/config-bgp-floating-ip-over-l2-segmented-network.html



More information about the openstack-discuss mailing list