Ubuntu Wiki

General Charm Updates

OpenStack Juno support for 14.04 and 14.10

All OpenStack charms now support deployment of OpenStack 2014.2 (Juno) on Ubuntu 14.04 LTS and Ubuntu 14.10; this support includes the following charms:

• keystone
• cinder
• glance
• nova-cloud-controller
• nova-compute
• quantum-gateway
• swift-proxy
• swift-storage
• ceilometer
• ceilometer-agent
• heat
• neutron-api
• neutron-openvswitch
• nova-cell

To deploy OpenStack Juno on Ubuntu 14.04, use the 'openstack-origin' configuration option, for example:

cat > config.yaml << EOF
nova-cloud-controller:
  openstack-origin: cloud:trusty-juno
EOF
juju deploy --config config.yaml nova-cloud-controller

OpenStack Juno is provided as the default OpenStack release on Ubuntu 14.10 so no additional configuration is required in 14.10 deployments.

Upgrading 14.04 deployments to Juno

WARNING: Upgrading an OpenStack deployment is always a non-trivial process. The OpenStack charms automate alot of the process, however always plan and test your upgrade prior to upgrading production OpenStack environments.

Existing Icehouse deployments of OpenStack on Ubuntu 14.04 can be upgraded to Juno by issuing:

juju upgrade-charm <charm-name>
juju set <charm-name> openstack-origin=cloud:trusty-juno

for each OpenStack charm in your deployment.

New Charm Features

Worker Thread Optimization

Where appropriate, the OpenStack charms will automatically configure appropriate worker values for API and RPC processes to optimize use of available CPU resources on deployed units. By default, this is set at twice the number of cores - however it can be tweaked using the worker-multiplier option provided by supporting charms:

juju set neutron-api worker-multiplier=4

the above example increases the default #cores x 2 to #cores x 4. Note that you should ensure that your mysql service can deal with the associated increase in concurrent connections - the mysql charm has a max-connections option that allows you to tweak this.

Network Segregation Configuration

The OpenStack charms feature support for use of multiple networks for separation of traffic; specifically:

• os-data-network: Data network for tenant network traffic supporting instances
• os-admin-network: Admin network - used for Admin endpoint binding and registration in keystone
• os-public-network: Public network - used for Public endpoint binding and registration in keystone

• os-internal-network: Internal network - used for internal communication between OpenStack services and for Internal endpoint registration in keystone

in addition the Ceph charms (ceph-osd, ceph) support splitting 'public' access traffic from 'cluster' admin and re-sync traffic, via the ceph-public-network and ceph-cluster-network configuration options.

All network configuration options should be provided in standard CIDR format - for example 10.20.0.0/16.

This feature should also support IPv6 networking as well, although this should be considered a technical preview for this release (see below).

IPv6 Support

NOTE: this feature only works as described under the Juno OpenStack release and should be considered a technical preview this cycle.

NOTE: this feature does not currently support IPv6 privacy extensions. In order for the charms to function correctly, privacy extensions must be disabled and a non-temporary address must be configured/available on your network interface.

A subset of the OpenStack charms now have a feature to prefer IPv6 networking for binding API endpoints and service-to-service communication:

• nova-cloud-controller
• nova-compute
• glance
• keystone
• ceph/ceph-osd
• neutron-api
• cinder
• openstack-dashboard

have been tested and are know to work in IPv6 configurations with the 'prefer-ipv6' configuration option enabled.

• swift-proxy
• swift-storage

also have this flag, but currently require a patched version of swift to function in an IPv6 environment. There are also changes proposed to the mysql, percona-cluster and rabbitmq-server charms which should land soon to enable this feature in other OpenStack supporting services.

Further enablement work will be done next cycle to complete this support across the charms, and hopefully have full upstream support for using IPv6 with OpenStack as well.

Neutron

The Neutron support in the OpenStack charms has been refactored into two new charms:

• neutron-api: Supporting API and central control operations.
• neutron-openvswitch: Supporting deployment of the Neutron ML2 plugin with Open vSwitch on nova-compute nodes.

These charms can be introduced into an existing OpenStack deployment:

juju deploy neutron-api
juju deploy neutron-openvswitch
juju add-relation neutron-api mysql
juju add-relation neutron-api keystone
juju add-relation neutron-api rabbitmq-server
juju add-relation neutron-api quantum-gateway
juju add-relation neutron-api neutron-openvswitch
juju add-relation neutron-api nova-cloud-controller
juju add-relation neutron-openvswitch rabbitmq-server
juju add-relation neutron-openvswitch nova-compute

Use of these two new charms also allows split of message brokers so that Nova and Neutron can use separate RabbitMQ deployments.

Use of these two new charms supports some additional features not enabled in the deprecated neutron support in nova-cloud-controller, specifically:

• Support for using the l2population driver for ARP table optimization at scale (l2-population configuration option - defaults to True).
• Support for using VXLAN overlay networks instead of GRE (overlay-network-type configuration option - defaults to GRE).

NOTE: the quantum-gateway charm now includes two amqp interfaces, one for neutron and one for nova, for use in split broker deployments. In single broker deployments, both neutron and nova will use the default 'amqp' relation for messaging configuration.

Nova Cells

The Nova charms now support deployment in Nova Cell configurations using the new nova-cell charm; See the nova-cell charm for details of how this works and how to use in a OpenStack deployment. A complete guide to this feature with example juju-deployer configurations will be posted soon.

Clustering

The hacluster charm has gone through some significant re-factoring to support changing configuration options post deployment, supporting upgrades of existing single network, clustered deployments to multi-network clustered deployments.

This charm also now supports direct configuration of the corosync bindiface and port in preference over any configuration provided from the principle charm its deployed with. Configuration of these options via the principle charm will be removed during the 15.04 cycle, users need to migrate to using the direct configuration options prior to the next stable release alongside 15.04.