ReleaseNotes1504

Differences between revisions 24 and 57 (spanning 33 versions)
⇤ ← Revision 24 as of 2015-01-30 11:18:01 →
Size: 7518
Editor: james-page
Comment:
← Revision 57 as of 2016-06-20 13:04:30 → ⇥
Size: 8627
Editor: james-page
Comment: Drop Ubuntu - that's kinda obvious
Deletions are marked like this. Additions are marked like this.
Line 1: Line 1:
## page was renamed from UbuntuOpenStack/OpenStackCharms/ReleaseNotes1504
## page was renamed from ServerTeam/OpenStackCharms/ReleaseNotes1504
## page was copied from ServerTeam/OpenStackCharms/ReleaseNotes1501
Line 5: Line 8:
The 15.01 OpenStack Charm release includes updates for the following charms: The 15.04 OpenStack Charm release includes updates for the following charms:
Line 26: Line 29:

This release has mainly been focussed on bug fixing, however some new features have been introduced.		  * percona-cluster

These release notes complement the [[ServerTeam/OpenStackCharms/ReleaseNotes1501|15.01]] charm release notes in terms of charm features delivered since the last Ubuntu release.
Line 31: Line 35:
=== Clustering ===

The trusty hacluster charm now supports running in multicast (default) and unicast modes, supporting use of this charm in environments where multicast UDP is not supported. To enable this feature:

{{{
juju set hacluster corosync_transport=unicast
}}}

At this point in time the previous node entries for the multicast cluster have to be removed manually - to complete this action, do the following on one of the members of the cluster:

{{{
sudo crm node list
}}}

New unicast nodes will start at 1001; the original multicast node entries should be deleted using:

{{{
sudo crm configure
 > delete <id>
}}}

The trusty hacluster charm now supports a new ‘debug’ configuration option to increase the verbosity of logging from corosync and pacemaker.

The trusty hacluster charm also includes a number of fixes to improve the way that quorum is handled by corosync and pacemaker.

=== Ceph ===

The ceph and ceph-osd charms now support setting sysctl options via charm configuration and provide a sensible default for the ‘kernel.pid_max’ sysctl option. This should support faster recovery in the event of a major outage in a Ceph deployment.

For charm authors, the ceph charm now has a Ceph broker API. This allows Ceph clients to request Ceph cluster actions e.g. create a new pool, via a new api as opposed to performing them on the client side. This will facilitate easily adding new functionality with reduced code impact while reducing the burden on the client to have to elect a leader to perform such actions, avoiding code duplication.

=== Ceph RADOS Gateway ===

The ceph-radosgw charm can now be deployed in a clustered configuration using a VIP as the object storage endpoint in conjunction with the hacluster charm:

{{{
juju deploy cs:trusty/hacluster hacluster-radosgw
juju deploy -n3 cs:trusty/ceph-radosgw
juju set ceph-radosgw vip=10.5.100.10
juju add-relation ceph-radosgw hacluster-radosgw
}}}

The ceph-radosgw charm now also support using an embedded webcontainer option provided natively by Ceph:

{{{
juju set ceph-radosgw use-embedded-webserver=true
}}}

This avoids using Apache2 and mod_fastcgi, which lacks support for chuck transfer encoding and 100-continue as provided in the Ubuntu archive.

=== Keystone ===

The keystone charm contains a number of improvements to support use of SSL endpoints in highly available deployments.

For charm authors, the keystone charm now has an additional ‘identity-notifications’ relation type; this relation is used by keystone to notify other charms when entries in the keystone service catalog change, and was introduced to support use with Ceilometer.

=== Glance ===

The glance charm now supports use of custom end-user provided configuration flags via the ‘config-flags’ charm option.

=== Ceilometer ===

The ceilometer charm can now be deployed in a clustered configuration using a VIP as the endpoint in conjunction with the hacluster charm:

{{{
juju deploy cs:trusty/hacluster hacluster-ceilometer
juju deploy -n3 cs:trusty/ceilometer
juju set ceilometer vip=10.5.100.20
juju add-relation ceilometer hacluster-ceilometer
}}}

The ceilometer charm now requires to relations to the keystone charm, to support both service catalog registration of the Ceilometer endpoint and notifications to changes to the service catalog:

{{{
juju add-relation ceilometer keystone:identity-service
juju add-relation ceilometer keystone:identity-notifications
}}}

NOTE: relation endpoint types must now be specified to avoid ambiguity.

=== Neutron ===

The quantum-gateway charm now has a fast failover option of neutron resources when multiple gateway units are used with the hacluster charm:

{{{
juju deploy cs:trusty/hacluster hacluster-ngateway
juju deploy -n3 cs:trusty/quantum-gateway neutron-gateway
juju set neutron-gateway vip=10.5.100.30
juju set neutron-gateway ha-legacy-mode=True
juju add-relation neutron-gateway hacluster-ngateway
}}}

NOTE: This feature has been introduced to support a level of resilience in Icehouse based deployments. Future charm work will include enablement of native Neutron support for router HA for later OpenStack releases.

The neutron charms now also support VLAN and flat networking in additional to GRE and VXLAN for tenant networks.

The quantum-gateway charm now also supports setting sysctl options via charm configuration inline with the ceph and nova-compute charms - see the ‘sysctl’ configuration option for details.

=== Swift ===

The swift-proxy charm now automatically rebalances rings on scale out of swift-storage services, allowing swift deployments to be managed more directly using Juju.

The swift-proxy charm also includes configuration options to set the minimum period between ring rebalancing and an option to completely disable rebalancing - this should be used when adding a number of new swift-storage service units to a deployment, to avoid re-balancing as every new set of storage is added to the swift deployment.

=== Nova ===

The nova-cloud-controller charm can now use memcache to store tokens for instance console access, supporting use of instance consoles in HA configuration via the OpenStack dashboard:

{{{
juju deploy -n3 cs:trusty/memcached
juju deploy -n3 cs:trusty/nova-cloud-controller
juju add-relation nova-cloud-controller memcached
}}}

The nova-cloud-controller charm now includes a ‘nova-alchemy-flags’ to allow Nova database configuration options to be directly tuned by charm users.

The nova-compute charm now has support for different storage backends; specifically it now supports local disk (default), Ceph RBD or LVM backends:

{{{
juju set nova-compute libvirt-image-backend=rbd
juju set nova-compute rbd-pool=nova
juju set nova-compute ceph-osd-replication-count=3
juju add-relation ceph nova-compute
}}}

The nova-compute charm now supports configuring disk cachemodes, allowing configuration of Nova disk usage options - see the ‘disk-cachemodes’ configuration option in the nova-compute charm and the upstream documentation about this feature.

The nova-compute charm now also supports setting sysctl options via charm configuration inline with the ceph and quantum-gateway charms - see the ‘sysctl’ configuration option for details.

=== OpenStack Dashboard ===

The openstack-dashboard charm now has improved support for use with multi-region clouds.

=== RabbitMQ ===

The rabbitmq-server charm now supports deployment in IPv6 only networks:

{{{
juju set rabbitmq-server prefer-ipv6=true
}}}

and use of a specific network for access to the message broker - see the ‘access-network’ configuration option for more details.		 === OpenStack Kilo support for 14.04 and 15.04 ===

All !OpenStack charms now support deployment of !OpenStack 2015.1 (Kilo) on Ubuntu 14.04 LTS and Ubuntu 15.04; this support includes the following charms:
 
 * keystone
 * cinder
 * cinder-ceph
 * glance
 * nova-cloud-controller
 * nova-compute
 * quantum-gateway
 * swift-proxy
 * swift-storage
 * ceilometer
 * ceilometer-agent
 * heat
 * neutron-api
 * neutron-openvswitch
 * openstack-dashboard

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

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

!OpenStack Kilo is provided as the default !OpenStack release on Ubuntu 15.04 so no additional configuration is required in 15.04 deployments.

'''NOTE''': In order to use Juju with Ubuntu 15.04, you'll have to use the 1.23 release which is currently in pre-release testing.

=== Neutron - Distributed Virtual Router ===

The neutron-api charm now has a enable-dvr option to turn on Distributed Virtual Routing. When this is enabled each Compute Node will provide North/South DNAT (floating IP traffic) as well as East/West L3 forwarding (traffic between VM in the same tenant), North/South SNAT still goes through the neutron gateway.

Prerequisites:
 * Compute having external connectivity
 * l2-population is enabled
 * !OpenStack release >= Juno
 * enable-l3ha is disabled
 * Overlay network type is vxlan

More details of DVR can be found [[https://wiki.openstack.org/wiki/Neutron/DVR|here]]

'''NOTE''': This feature was considered experimental in !OpenStack Juno so is only recommended for use with !OpenStack Kilo.

=== Neutron - Router High Availability ===

The neutron-api charm now has a enable-l3ha option. If this option is enabled then subsequent routers are created with the 'ha' flag set. This will enable automatic failover of a router should the gateway unit that it resides on fail.

Prerequisites:
 * More than 1 neutron gateway node
 * l2-population is disabled
 * !OpenStack release >= Juno
 * enable-dvr is disabled
 * Overlay network type is vlan, gre or vxlan

More details on L3 HA can be found [[https://wiki.openstack.org/wiki/Neutron/L3_High_Availability_VRRP|Here]]

'''NOTE''': This feature was considered experimental in !OpenStack Juno so is only recommended for use with !OpenStack Kilo.

=== Percona Cluster - Improved HA failure detection ===

The percona-cluster charm now ensures that the access Virtual IP is always co-located with an active, up-to-date instance of Percona XtraDB Cluster.

=== ZeroMQ ===

The Nova and Neutron charms now provide initial support use of ZeroMQ as an alternative to RabbitMQ for messaging; this is only supported with OpenStack Kilo or later.

To deploy Nova and Neutron with ZeroMQ, make use of the new openstack-zeromq charm (sample subset of juju commands):

{{{
juju deploy redis-server
juju deploy openstack-zeromq
juju add-relation redis-server openstack-zeromq
for svc in nova-cloud-controller nova-compute neutron-api neutron-openvswitch quantum-gateway; do
   juju deploy $svc
   juju add-relation $svc openstack-zeromq
done
}}}

At this point in time ZeroMQ should be considered for testing only. Further work is expected both upstream in !OpenStack and across the !OpenStack charm set during the next development cycle.

=== Deployment from source ===

'''EXPERIMENTAL FEATURE'''

The following charms now have support for deploying OpenStack directly from git repositories:

 * cinder
 * keystone
 * glance
 * nova-cloud-controller
 * nova-compute
 * neutron-api
 * quantum-gateway
 * neutron-openvswitch

This feature is enabled by using the new 'openstack-origin-git' configuration option; this can be used to configure a minimal or expanded list of repositories to collate and install - see the README in each charm for specific details.

Further deploy from source enablement is expected during the 15.07 charm development cycle.

'''NOTE''': Current support extends to deployment of Icehouse and Juno from source repositories; Kilo support is not yet complete but will be delivered as a stable update between now and the 15.07 charm release.

=== Monitoring ===

The Openstack charms can now be related to the nrpe charm to provide basic Nagios monitoring. To allow alert thresholds to be tweaked per service deploy each a nrpe subordinate charm per Openstack Service. e.g.

{{{
juju deploy nrpe nrpe-glance
juju deploy nrpe nrpe-cinder
juju deploy nagios
juju deploy glance
juju deploy cinder
juju add-relation nrpe-glance glance
juju add-relation nrpe-glance nagios
juju add-relation nrpe-cinder cinder
juju add-relation nrpe-cinder nagios
}}}

The thresholds can then tweaked:

{{{
juju set nrpe_glance "load=-w 2,2,2 -c 4,4,4"
juju set nrpe_cinder "load=-w 4,4,4 -c 10,10,10"
}}}

=== SSL/HTTPS/PKI ===

All service endpoints can now be configured as https either by having the Keystone charm act as a certificate authority and auto-generate cert/key for each endpoint or by injecting cert/key/ca to be able to use an existing CA. Endpoints can also now choose to use PKI tokens in keystone although the default remains UUID.

To enable https endpoints (using keystone as CA):

{{{
juju set keystone https-service-endpoints='true'
}}}

To enable PKI tokens:

{{{
juju set keystone enable-pki='true'
}}}

For more ssl options see config.yaml

== Deprecation Notices ==

=== quantum-gateway ===

15.04 is the last release of the 'quantum-gateway' charm; for 15.07 it will be renamed 'neutron-gateway'. Migration instructions will be provided for 15.07 release.

=== MySQL HA with Ceph ===

The original HA implementation of OpenStack used the MySQL charm in-conjunction with Ceph to provide a traditional active/passive shared block device HA mysql option.

Percona XtraDB Cluster, as provided by the percona-cluster charm, provides an active/active MySQL configuration with no requirement for a shared block device.

Users of the original MySQL HA option should migrate to a Percona XtraDB Cluster deployment between now and the 15.10 charm release, where support for MySQL with Ceph will be removed.

=== hacluster - default transport ===

For the 15.04 charm release, the hacluster charm will continue to use multicast as its default transport option for cluster communications.

Unicast is generally more universally usable across all Juju providers - we'll be switching the default for 15.07 including automatic migration and re-configuration of existing multicast clusters. Instructions will also be provided for users who wish to continue to use the multicast transport option.

=== nova-cloud-controller and neutron-api charms ===

Management of neutron configuration and services will be removed from any nova-* charms for the 15.10 charm release.

This means that the neutron-api principle charm needs to be deployed to run the neutron-server service and if openvswitch is being used then neutron-openvswitch subordinate charm needs to deployed and attached to the nova-compute charm. This is a fully supported migration path for existing deployments not using these charms.

The nova-compute charm now has an option called 'manage-neutron-plugin-legacy-mode'. The default value for this option is True but as of 15.07 this will switch to False which will disable the management of neutron services by default.
Line 176: Line 213:
For the full list of bugs resolved for the 15.01 release please refer to https://launchpad.net/charms/+milestone/15.01 For the full list of bugs resolved for the 15.04 release please refer to https://launchpad.net/charms/+milestone/15.04

Contents

  1. Summary
  2. New Charm Features
    1. OpenStack Kilo support for 14.04 and 15.04
    2. Neutron - Distributed Virtual Router
    3. Neutron - Router High Availability
    4. Percona Cluster - Improved HA failure detection
    5. ZeroMQ
    6. Deployment from source
    7. Monitoring
    8. SSL/HTTPS/PKI
  3. Deprecation Notices
    1. quantum-gateway
    2. MySQL HA with Ceph
    3. hacluster - default transport
    4. nova-cloud-controller and neutron-api charms
  4. Bugs Fixed

Summary

The 15.04 OpenStack Charm release includes updates for the following charms:

  • ceilometer
  • ceilometer-agent
  • ceph
  • ceph-radosgw
  • cinder
  • cinder-ceph
  • glance
  • hacluster
  • heat
  • keystone
  • neutron-api
  • neutron-openvswitch
  • nova-cloud-controller
  • nova-compute
  • openstack-dashboard
  • quantum-gateway
  • rabbitmq-server
  • swift-proxy
  • swift-storage
  • percona-cluster

These release notes complement the 15.01 charm release notes in terms of charm features delivered since the last Ubuntu release.

New Charm Features

OpenStack Kilo support for 14.04 and 15.04

All OpenStack charms now support deployment of OpenStack 2015.1 (Kilo) on Ubuntu 14.04 LTS and Ubuntu 15.04; this support includes the following charms:

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

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

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

OpenStack Kilo is provided as the default OpenStack release on Ubuntu 15.04 so no additional configuration is required in 15.04 deployments.

NOTE: In order to use Juju with Ubuntu 15.04, you'll have to use the 1.23 release which is currently in pre-release testing.

Neutron - Distributed Virtual Router

The neutron-api charm now has a enable-dvr option to turn on Distributed Virtual Routing. When this is enabled each Compute Node will provide North/South DNAT (floating IP traffic) as well as East/West L3 forwarding (traffic between VM in the same tenant), North/South SNAT still goes through the neutron gateway.

Prerequisites:

  • Compute having external connectivity
  • l2-population is enabled

  • OpenStack release >= Juno

  • enable-l3ha is disabled
  • Overlay network type is vxlan

More details of DVR can be found here

NOTE: This feature was considered experimental in OpenStack Juno so is only recommended for use with OpenStack Kilo.

Neutron - Router High Availability

The neutron-api charm now has a enable-l3ha option. If this option is enabled then subsequent routers are created with the 'ha' flag set. This will enable automatic failover of a router should the gateway unit that it resides on fail.

Prerequisites:

  • More than 1 neutron gateway node
  • l2-population is disabled

  • OpenStack release >= Juno

  • enable-dvr is disabled
  • Overlay network type is vlan, gre or vxlan

More details on L3 HA can be found Here

NOTE: This feature was considered experimental in OpenStack Juno so is only recommended for use with OpenStack Kilo.

Percona Cluster - Improved HA failure detection

The percona-cluster charm now ensures that the access Virtual IP is always co-located with an active, up-to-date instance of Percona XtraDB Cluster.

ZeroMQ

The Nova and Neutron charms now provide initial support use of ZeroMQ as an alternative to RabbitMQ for messaging; this is only supported with OpenStack Kilo or later.

To deploy Nova and Neutron with ZeroMQ, make use of the new openstack-zeromq charm (sample subset of juju commands): 

juju deploy redis-server
juju deploy openstack-zeromq
juju add-relation redis-server openstack-zeromq
for svc in nova-cloud-controller nova-compute neutron-api neutron-openvswitch quantum-gateway; do
   juju deploy $svc
   juju add-relation $svc openstack-zeromq
done

At this point in time ZeroMQ should be considered for testing only. Further work is expected both upstream in OpenStack and across the OpenStack charm set during the next development cycle.

Deployment from source

EXPERIMENTAL FEATURE

The following charms now have support for deploying OpenStack directly from git repositories:

  • cinder
  • keystone
  • glance
  • nova-cloud-controller
  • nova-compute
  • neutron-api
  • quantum-gateway
  • neutron-openvswitch

This feature is enabled by using the new 'openstack-origin-git' configuration option; this can be used to configure a minimal or expanded list of repositories to collate and install - see the README in each charm for specific details.

Further deploy from source enablement is expected during the 15.07 charm development cycle.

NOTE: Current support extends to deployment of Icehouse and Juno from source repositories; Kilo support is not yet complete but will be delivered as a stable update between now and the 15.07 charm release.

Monitoring

The Openstack charms can now be related to the nrpe charm to provide basic Nagios monitoring. To allow alert thresholds to be tweaked per service deploy each a nrpe subordinate charm per Openstack Service. e.g. 

juju deploy nrpe nrpe-glance
juju deploy nrpe nrpe-cinder
juju deploy nagios
juju deploy glance
juju deploy cinder
juju add-relation nrpe-glance glance
juju add-relation nrpe-glance nagios
juju add-relation nrpe-cinder cinder
juju add-relation nrpe-cinder nagios

The thresholds can then tweaked: 

juju set nrpe_glance "load=-w 2,2,2 -c 4,4,4"
juju set nrpe_cinder "load=-w 4,4,4 -c 10,10,10"

SSL/HTTPS/PKI

All service endpoints can now be configured as https either by having the Keystone charm act as a certificate authority and auto-generate cert/key for each endpoint or by injecting cert/key/ca to be able to use an existing CA. Endpoints can also now choose to use PKI tokens in keystone although the default remains UUID.

To enable https endpoints (using keystone as CA): 

juju set keystone https-service-endpoints='true'

To enable PKI tokens: 

juju set keystone enable-pki='true'

For more ssl options see config.yaml

Deprecation Notices

quantum-gateway

15.04 is the last release of the 'quantum-gateway' charm; for 15.07 it will be renamed 'neutron-gateway'. Migration instructions will be provided for 15.07 release.

MySQL HA with Ceph

The original HA implementation of OpenStack used the MySQL charm in-conjunction with Ceph to provide a traditional active/passive shared block device HA mysql option.

Percona XtraDB Cluster, as provided by the percona-cluster charm, provides an active/active MySQL configuration with no requirement for a shared block device.

Users of the original MySQL HA option should migrate to a Percona XtraDB Cluster deployment between now and the 15.10 charm release, where support for MySQL with Ceph will be removed.

hacluster - default transport

For the 15.04 charm release, the hacluster charm will continue to use multicast as its default transport option for cluster communications.

Unicast is generally more universally usable across all Juju providers - we'll be switching the default for 15.07 including automatic migration and re-configuration of existing multicast clusters. Instructions will also be provided for users who wish to continue to use the multicast transport option.

nova-cloud-controller and neutron-api charms

Management of neutron configuration and services will be removed from any nova-* charms for the 15.10 charm release.

This means that the neutron-api principle charm needs to be deployed to run the neutron-server service and if openvswitch is being used then neutron-openvswitch subordinate charm needs to deployed and attached to the nova-compute charm. This is a fully supported migration path for existing deployments not using these charms.

The nova-compute charm now has an option called 'manage-neutron-plugin-legacy-mode'. The default value for this option is True but as of 15.07 this will switch to False which will disable the management of neutron services by default.

Bugs Fixed

For the full list of bugs resolved for the 15.04 release please refer to https://launchpad.net/charms/+milestone/15.04

OpenStack/OpenStackCharms/ReleaseNotes1504 (last edited 2016-06-20 13:04:30 by james-page)