The 17.08 OpenStack Charm release includes updates for the following charms:
New Charm Features¶
OpenStack Pike support on 16.04¶
The charms provide full support for OpenStack Pike. For further details and documentation on Openstack Pike please check out https://releases.openstack.org/pike
To deploy OpenStack Pike on Ubuntu 16.04, use the ‘openstack-origin’ configuration option, for example:
cat > config.yaml << EOF nova-cloud-controller: openstack-origin: cloud:xenial-pike EOF juju deploy --config config.yaml nova-cloud-controller
To upgrade an existing Ocata based deployment on Ubuntu 16.04 to the Pike release, re-configure the charm with a new openstack-origin configuration:
juju config nova-cloud-controller openstack-origin=cloud:xenial-pike
Please ensure that ceph services are upgraded before services that consume ceph resources, such as cinder, glance and nova-compute:
juju config ceph-mon source=cloud:xenial-pike juju config ceph-osd source=cloud:xenial-pike
Upgrading an OpenStack cloud is still not without risk; upgrades should be tested in pre-production testing environments prior to production deployment upgrades.
The OpenStack charms now support deployment of Gnocchi for storage of resources, metrics and associated measures generated by Ceilometer from across an OpenStack Cloud.
This preview release provides support for use with Ceph for measure storage, with MySQL providing the supporting database infrastructure. The first GA release of the Gnocchi charm will be made independently of the next charm release in October.
To deploy the preview version of the Gnocchi charm:
juju deploy cs:~openstack-charmers-next/gnocchi juju deploy cs:memcached juju add-relation gnocchi ceph-mon juju add-relation gnocchi ceilometer juju add-relation gnocchi mysql juju add-relation gnocchi keystone juju add-relation gnocchi memcached
The Ceilometer API will be automatically disabled as soon as Ceilometer is configured to store cloud telemetry data in Gnocchi.
The Gnocchi charm only supports deployment of Gnocchi with OpenStack Ocata or later.
Neutron Internal DNS¶
The Neutron charms now support use of the Neutron ML2 DNS extension driver, supporting management of DNS records for instances internally to Neutron.
To enable this feature please review the ‘enable-ml2-dns’ and ‘dns-domain’ options of the neutron-api charm, and the ‘dns-servers’ option of the neutron-gateway and neutron-openvswitch charms.
Improved application configuration in LXD deployments¶
Deployment of charms providing API services in LXD containers now calculates worker processes more intelligently, capping at 4 worker processes unless an explicit value is provided via the ‘worker-multiplier’ option. This limits default memory consumption in converged architectures where API services may be placed in densely packed LXD containers on high specification hardware.
The percona-cluster charm also now limit the InnoDB buffer pool size to a maximum of 512MB (or 50% of system memory on smaller deployments) unless explicit configuration is provided to the charm.
Percona Cluster password management¶
Previous versions of the percona-cluster charm required that the SST and root passwords be provided via configuration options set prior to deployment.
If the ‘root-password’ and ‘sst-password’ options are not provided, the charm will now generate passwords automatically across a clustered multi-unit deployment.
Percona Cluster tuning¶
The percona-cluster charm now provides a high-level ‘tuning-level’ option, supporting three configuration levels:
- safest (default): use configuration options with best data
- fast: compromise some data integrity guarantees
to improve performance.
- unsafe: pretty much throw away all data integrity
guarantees to maximise performance.
Use of the ‘fast’ option may be appropriate in multi-unit deployments where some level of consistency can be sacrificed at the individual unit level due to the clustered nature of the deployment.
DVR + Layer 3 HA¶
The neutron-api charm now supports enabling DVR and Layer 3 HA at the same time for OpenStack Newton clouds or later.
Ceph Bluestore support¶
The Ceph charms now support the latest Luminous stable release, including the new Bluestore storage backend for OSD devices. This is not currently the default for charm based Ceph deployments; Bluestore may become the default in a later charm release pending further performance and resilience testing.
To enable use of Bluestore for OSD devices:
juju config ceph-osd bluestore=true
Please ensure that the keystone charm is upgraded first.
To upgrade an existing deployment to the latest charm version simply use the ‘upgrade-charm’ command:
juju upgrade-charm cinder
ZeroMQ messaging support across the charms will be removed during the Queens development cycle.
PostgreSQL database support across the charms will be removed during the Queens developement cycle.
Deploy from Source (DFS) support is under review for sustainability and may be removed during the Queens development cycle.
By default, the hacluster charm will assume a cluster size of three units; in order to scale a three unit clustered application out, you must increase the cluster-count configuration option to the target size prior to adding the additional units to the clustered application.
Liberty to Mitaka upgrades¶
When upgrading Trusty/Liberty clouds to Trusty/Mitaka using the configuration option ‘big bang’ approach to upgrading, the neutron-gateway charm incorrectly configures the neutron-openvswitch-agent due to a configuration file name change. This can be resolved post upgrade completion by triggering execution of the config-changed hook by toggling debug logging on and off.
This problem does not occur when action managed upgrades are used to upgrade the neutron-gateway application.
heat encryption keys¶
In HA deployments individual units of the heat application are configured with different encryption keys resulting in issues when authentication information is stored by units in the database. This bug will be fixed in a stable charm update after the 17.08 release.
Nova LXD Ceph support¶
Support for presentation of Ceph backed Cinder volumes to Nova LXD instances is supported in the Pike OpenStack release. This integration uses the Linux kernel RBD module and requires a minimum of a Linux 4.5 to support the default crush tunables for Luminous.
Deployments using the release Xenial kernel (4.4) should set the crush tunables level to ‘hammer’:
juju run --unit ceph-mon/0 'sudo ceph osd crush tunables hammer'
Xenial deployments using hardware enablement kernels (hwe) or deployments on later Ubuntu releases do not require this workaround.
For the full list of bugs resolved for the 17.08 release please refer to https://launchpad.net/openstack-charms/+milestone/17.08