⁠Chapter 2. Release Information

Copy-on-Write cloning for RBD-backed disks is now supported. As such, the Compute service no longer needs to download a glance image to local disk and then copy it again to RBD. Rather, disks can now be efficiently created directly in the Ceph system, thereby speeding up the creation of instances and, in the process, saving bandwidth and disk space.

An admin user can now specify the Provider network type (the physical mechanism by which the virtual network is implemented), when creating a new network. Previously, the dashboard (horizon) defaulted to the 'Local' provider network type, and it was not possible to select another type. The types 'Flat', 'VLAN', 'GRE', and 'VXLAN', and 'Local' can now be selected in the new 'Provider Network Type' drop-down field. Depending on the type, a segmentation ID, tunnel ID, or physical network name must be additionally specified.

This feature exposes interactive web-based serial consoles to openstack VMs through a websocket proxy. Generally used as a debugging tool (for example, VMs can be accessed even if network configuration fails).

A new service (websocket proxy) is now available that handles websocket connections to the serial consoles of the VMs. The websocket proxy can be deployed on a machine other than from the hypervisor.

The host argument for the 'nova evacuate' command has been made optional. This means that the user no longer has to know the host destination, simplifying evacuation in the case of an unplanned failure.

This enhancement enables changes to a subnet's IP address allocation pool using the update command. Previously, administrators were unable to change the allocation pool range for a subnet.

If shrinking the pool, consideration must be given to IP addresses that have already been allocated.

Compute now automatically attempts a controlled shutdown for stop, rescue, and delete instance actions. If the controlled shutdown fails, Compute falls back to a forced shutdown.

Previously, when a rescue was done on an instance, the image_base_ref in the instance_meta_data was used. If this attribute was not populated, the custom image_ref was used instead. This posed a problem where the custom image used might be corrupt, leading to errors; or too large, leading to timeouts. Also, if the base image was deleted, the image ref on the instance_system_metadata would be invalid, leading to the rescue operation failing. With this release, the user can now specify which image is to be used for rescue (this could be a default base image, or a custom image).

With this feature, there is now a standardized data reporting mechanism for VM diagnostics. Previously, the diagnostics API had no formal specification. As a result, different virtualization drivers in the Compute service reported a different data set and made it difficult for VM diagnostics to consume the data in a predictable manner. With this update, a new version of the diagnostic API was implemented in the version 3 API, and it now defines a standardized set of data items that virtualization drivers must use for reporting.

OpenStack Compute now supports associating SR-IOV PCI devices with networks and binding OpenStack Networking SR-IOV ports to them. PCI-Passthrough to SR-IOV virtual functions provide direct access to networking hardware specialized for virtualization with one physical device supporting multiple virtual machines. By supporting SR-IOV devices, virtual machines can now employ SR-IOV hardware for networking.

Role-based access control (RBAC) checks are now supported for actions that interact with the Compute service (nova); rules are defined in the /etc/openstack-dashboard/nova_policy.json configuration file. RBAC checks allow an administrator to finely tune a user's access. For example, an administrator might allow end users to view the complete flavor listing.

Role-based access control (RBAC) checks are now supported for actions calling the network service; rules are defined in the /etc/openstack-dashboard/neutron_policy.json configuration file. RBAC checks allow an administrator to finely tune a user's access. For example, an administrator might prevent end users from creating a subnet or changing a firewall policy.

An admin user can now evacuate a compute host using the dashboard. Two tabs now provide information for hypervisors: 'Hypervisor' and 'Compute Host' (Admin > Hypervisors). If a host is down, an 'Evacuate Host' action is now visible for it on the Compute Host tab (providing a modal window to perform the evacuation).

Support has been added for Block Storage volume backups in the dashboard. Users can now create, view, delete, and restore volume backups. Note: This functionality is not displayed by default. To display volume-backup action items, update the /etc/openstack-dashboard/local_settings file with:

OPENSTACK_CINDER_FEATURES = { 'enable_backup': True, }

After updating the file, restart the httpd service with 'systemctl restart httpd'.

OpenStack Networking (neutron)-related features (for example, LBaaS, FWaaS, or VPNaaS) are now disabled or enabled using OpenStack Networking's extension list. OpenStack Networking feature settings have been removed from the dashboard's local_settings file.

Additionally, the default settings of enable_lb, firewall, and vpn have been changed to 'True'. Previously, the default was 'False' to avoid user confusion because LBaaS, FWaaS, and VPNaaS are optional features in OpenStack Networking. With this change, these features in the dashboard are enabled dynamically, so the 'True' default is more reasonable. (FWaaS and VPNaaS are currently included in this release as technology preview.)

An additional 'Action Log' tab is now available for specific instances (Project > Compute > Instances > [instance]). The tab lists all actions which have been carried out on that specific instance. For example, a tenant user can now use the 'Action Log' tab to see who created or shut down an instance.

With this feature, there is now a widget for managing the Image service metadata dictionary. The admin user is now able to edit properties of images directly under admin/images/edit.

When using OpenStack Networking (neutron) with the dhcp_agent_scheduler extension, it is now possible to add and remove DHCP agents from networks using the dashboard. This makes it easier to manage the high availability of DHCP agents for OpenStack Networking. When logged in as admin and navigating to the Admin Networks panel, a new DHCP Agents column with the number of agents associated with each network is now visible. Clicking on a network name displays the network's details together with a new 'DHCP Agents' table where the admin can add and delete agents.

Previously, there was no interface to allow assigning domain role to users. In this release, the Dashboard has been extended to support managing roles and users in different domains.

The Orchestration service now allows the user to update a stack in a FAILED state. Previously, failed stacks could only be deleted, not updated.

The Telemetry service now uses the newer oslo.messaging library for RPC, thereby replacing the deprecated openstack.common.rpc library.

This enhancement adds high availability for OpenStack Networking (neutron) virtual routers. This was added due to the impact of virtual routers going down with a network node; instances would lose external connectivity. Virtual routers can now be created with the 'High availability' flag, if the administrator sets it as the default. As a result, routers will then be created on multiple network nodes, with a designated single active instance node. The active node forwards traffic while the standbys monitor the master. In the event of failure impacting the active node, one of the standby will take over as the new active node.

With this feature, you can now create and deploy Distributed Virtual Routers (DVR) using the Open vSwitch (OVS) infrastructure (DVR is currently included as a technology preview feature).

Tenant networks can now be created that use the 'dnsmasq' process inside the DHCP agent to serve additional configuration to IPv6 DHCP clients, including addressing and support for IPv6 stateful subnets. Note that the default gateway is still set by Router Advertisement messages sent by the L3 agent.

IMPORTANT: This feature requires dnsmasq-2.66-13.el7_1 or later.

With this enhancement, Tenant networks can now be created that use the 'dnsmasq' process inside the DHCP agent to serve additional configuration to IPv6 DHCP clients, including support for IPv6 stateless subnets.

This enhancement allows the creation of Tenant networks that use the 'radvd' process within the L3 agent for Router Advertisement messages. As a result, instances are able to use Stateless Address Autoconfiguration (SLAAC) or DHCPv6 to configure their IPv6 networking.

Extra Specs support for volume types has been added to the dashboard. An admin can now add additional keys and values to volume types (GUI implementation of the 'cinder type-key' command). To view extra specs, select Admin> Volumes > Volume Types, and click the type's 'View Extra Specs' action.

The code in the glance.store package used to store images into different store backends has been pulled out into a self-standing library.

OpenStack Networking (neutron) has introduced new attributes for IPv6 networks: 'Router Advertisement' and 'Address Assignment', which enables IPv6 subnets to be configured with more granularity. If OpenStack Networking is in use and an IPv6 subnet is being created, the dashboard now offers the following options in the 'IPv6 Address Configuration Mode' drop-down field:

--"SLAAC: Address discovered from Openstack Router"

--"DHCPv6 stateful: Address discovered from Openstack DHCP"

--"DHCPv6 stateless: Address discovered from Openstack Router and info from Openstack DHCP"

Providing no option means that addresses are configured manually or by a non-OpenStack system.

The ability for an administrator to manage image metadata (custom properties) has been added to the dashboard. The admin user can now add, update, or delete image metadata (implements the 'glance image-update <imageID> --property <key>=<value>' command). To view or update an image's metadata, select Admin > System > Images, and click the image's 'Update Metadata' action.

Role-based access control (RBAC) checks are now supported for actions that interact with the Orchestration service (heat); rules are defined in the /etc/openstack-dashboard/heat_policy.json configuration file. RBAC checks allow an administrator to finely tune a user's access. For example, an administrator might prevent end users from changing a stack template.

In this release, the following basic support for OpenStack Database-as-a-Service (trove) has been added to the dashboard:

* Datastore type/version drop down in Launch Instance

* Datastore type/version in Instance Details.

With this feature, there is now a separate Identity dashboard.

The Telemetry service now supports CSV download of daily usage data. In the dashboard, this feature is available through the 'Download CSV summary' button of the Admin > Overview panel. This button also becomes available after querying usage data.

The Identity service now supports domain-specific identity backends. This allows for a single Identity service instance to use multiple identity backends, such as centralized LDAP for normal users and a Identity-specific SQL database for service users.

MongoDB is still the preferred Telemetry back end for large installations; however, with this release SQL now features improved read/write processing, which may be suitable for small installations. This was accomplished through extensive refactoring of the database tables, query structuring, and optional parallelization of the ceilometer-collector service.

The user can now sort tables by timestamp in the dashboard (a timestamp parser has been added). For example, in the Project > Compute > Overview window, the user can now sort instances by 'Time since created'.

Table filtering has been updated in the dashboard to use API query attributes. A drop-down box and an input field for filtering have been added to tables for admin instances, admin images, and project instances. For example, the admin instances table might be filtered for 'Status=Active'.

With this release, parameters in HOT templates can now be defined with the type "boolean".

Code for the OpenStack Data Processing (sahara) dashboard has been merged into the OpenStack dashboard (horizon) code. If OpenStack Data Processing is correctly installed (openstack-sahara) and configured, no further dashboard configuration is necessary to display the 'Data Processing' tab for each region (Project > Data Processing).

This feature improves the workflow of creating a load balancer. You can now create a load balancer by specifying the IP address and port number.

You can now configure the openstack-cinder-volume service to limit I/O bandwidth when copying data between volumes. This allows you to throttle the load placed on the openstack-cinder-volume node for long-running I/O copy operations. To use this feature, configure the 'volume_copy_blkio_cgroup_name' and 'volume_copy_bps_limit' settings in /etc/cinder/cinder.conf of the openstack-cinder-volume node accordingly.

The state of an instance provided potentially interesting metrics, particularly to consumers of polled instance-related metrics. These metrics were already available to pollsters, but was not included in the resource metadata recorded for the instance in previous releases.

With this release, the current instance state is now included in resource metadata reported for polled instance-related meters.

The systemd script no longer explicitly sets a log file. This ensures that syslog settings take precedence during logging.

A new feature allows users to download an image's data partially and restart the download at any time, as long as the image data is available in the server. This has been implemented using the 'Content-Range' header, which follows the form "Content-Range: bytes FIRST_BYTE-LAST_BYTE/INSTANCE_BYTES". This header will be parsed only on download requests and there is no support on the client library yet.

Refer to the HTTP's specification for more information about the specific forms of this header.

The Orchestration service now supports an OS::Nova::ServerGroup resource, which allows you to apply scheduling constraints (like affinity or anti-affinity) to a group of servers.

This enhancement enables ipset kernel groups to be used for matching IP addresses in iptables security groups. The previous implementation of security groups, which made intensive use of iptables rules, resulted in an exponential growth of iptables rules in some cases. Specifically, multiple IP addresses previously needed to be added to the security groups of each Compute node's network port. As a result of this enhancement, the size of iptables rules on Compute nodes are significantly reduced, resulting in a performance increase in accepting new connections.

Previously, the Telemetry service always aggregated across all disks associated with an individual instance, making it impossible to bill for individual volume. This release now features the ability to gather disk metrics for individual devices. Specifically, the Telemetry service now supports new per-device disk.device.{read|write}.{requests|bytes} meters, in addition to the original aggregated disk.{read|write}.{requests|bytes} meters.

Orchestration now supports OS::Sahara::Cluster, OS::Sahara::NodeGroupTemplate, and OS::Sahara:ClusterTemplate resource types, so that users can manage OpenStack Data Processing resources through Orchestration templates.

The Telemetry service can now poll for SNMP metrics in a generic way. This allows the Telemetry service to gather additional SNMP metricts without writing a new pollster for each new metric.

OpenStack Compute can now optionally provide a config drive to instances based on a property on the image in the OpenStack Image service. Previously, Compute configuration determined whether a config drive was used and what format to use for it. With this update, users can now indicate config drive requirements using image properties.

A new rule now makes it possible to restrict an image's downloads by using the built-in policy engine. This rule ("download_image": "role:admin or rule:restricted") must be added to the policy.json file to ensure it is effective.

The Orchestration service now features an OS::Glance::Image resource type, which allows images to be created in the Image service as part of an Orchestration template.

A new OS-ENDPOINT-POLICY extension has been added to the Identity service. This extension allows for assigning separate policies to specific services or even individual endpoints. Policy assignment is now more flexible.

A 'Metadata' column has been added to the Flavors table (Admin > System > Flavors) that displays whether extra specs have been specified for a flavor ('Yes' or 'No'). The user can now click on either the column value or the 'Update Metadata' action to view or update defined metadata.

With this release, floating IP addresses created in Orchestration templates can now be updated in-place to connect to a different server. Previously a change to a template to move the attached port would result in a new Floating IP being created and the existing one removed.

In previous releases, every virtual CPU was configured as a socket. Some guest operating systems have arbitrary limits on the number of sockets they support, but are not limited in the number of cores or threads. This prevented an instance's OS from taking full advantage of the virtual CPUs configured.

With this release, the Compute service can now control an instance's virtual CPU topology. This allows an administrator and/or tenant users to specify constraints for the number of threads, cores and sockets to use for a guest instance. The Compute service will use the constraint information to configure a suitable guest CPU topology. With this, a guest OS such as Windows can take full advantage of all virtual CPUs without encountering support limits.

Administrators can now reset the state of a volume in the dashboard. This change permits an operator to select any valid status, regardless of the current status of the volume. Previously, this functionality was only available through the CLI command: # cinder reset-state --state available <volume-id>

Previous Compute versions delegated all CPU placement to the operating system kernel. Although the kernel attempted to keep guest processes running on a single NUMA node, this was not enforced. This meant that guests could drift across NUMA nodes, resulting in an inefficient usage of host resources and limiting guest performance.

With this update, Compute can now place guest instances on specific host NUMA nodes. The cloud administrator or tenant user can set preferences for the guest NUMA topology layout by enabling a scheduler filter that performs intelligent NUMA placement (affinity server group using hw:numa_policy=strict metadata). Compute takes into account the guest topology and then pins the guest instance to one or more host NUMA nodes, resulting in a more consistent guest performance and efficient use of host resources.

Administrators can now reset the state of a snapshot in the dashboard. Previously, this functionality was only available through the CLI command: # cinder snapshot-reset-state --state available <snapshot-id>

With this release, the Orchestration service (heat) now features the OS::Heat::SwiftSignal and OS::Heat::SwiftSignalHandle resource types, which provide an alternative implementation of WaitConditions that use OpenStack Object Storage (swift) to store data. The new types allow a user to wait for a server's configuration to finish before proceeding with the creation of other resources in the stack, and to optionally pass data from that server to another stack resource.

You can now configure a separate storage back end for Telemetry alarms. In previous releases, alarm-related data could only be stored in the same storage back-end as sample data. This was inconvenient, as neither the volume nor the nature of these data are similar. On a different back end, however, different criteria can now be usefully applied when selecting the appropriate database.

Support has been added to the dashboard for OpenStack Database-as-a-Service (trove). The management of incremental backups is now supported.

With this enhancement, all tables are now included during the creation of the database schema. This behavior allows for easier plugin management. Consequently, all OpenStack Networking (neutron) tables are present in the database after upgrading to Red Hat Enterprise Linux OpenStack Platform 6.

You can now perform a 'cinder retype' through the dashboard. This allows you to migrate volumes or to change any volumes setting (that are set from the volume's type) through the web interface.

A single guest can now have multiple network interfaces attached to the same logical host network. Previous versions of OpenStack Compute had an artificial restriction that a single guest cannot have multiple network interfaces connected to the same host network. There are, however, some valid use cases where this is required and thus Compute could not satisfy those use cases. With this update, the tenant user can now set up guest network interfaces without any restrictions imposed by Compute.

Support has been added for intelligent NUMA node placement for guests that have been assigned a host PCI device. PCI I/O devices, such as Network Interface Cards (NICs), can be more closely associated with one processor than another. This is important because there are different memory performance and latency characteristics when accessing memory directly attached to one processor than when accessing memory directly attached to another processor in the same server. With this update, Openstack guest placement can be optimized by ensuring that a guest bound to a PCI device is scheduled to run on a NUMA node that is associated with the guest's pCPU and memory allocation. For example, if a guest's resource requirements fit in a single NUMA node, all guest resources will now be associated with the same NUMA node.

The MongoDB back end can now persist notification payloads, which in turn allows event processing for Telemetry services configured to use a MongoDB back end. In previous releases, this functionality was only available through the SQLAlchemy back end, which was not fully supported in Red Hat Enterprise Linux OpenStack Platform.

Functionality for Block Storage Quality of Service (QoS) extra specs management such as maximum IO/seconds (maxIOPS) is now available in the administrator dashboard. Previously, QoS specs and associations had to be managed using cinder CLI commands (for example, cinder-qos-create, cinder-qos-delete, cinder-qos-key, or cinder-qos-associate).

With this release, you can perform most of these functions through the dashboard. To do so, select Admin > Volumes > Volume Types. From there, you can view, create, associate, or otherwise manage QoS specs.

It is now possible to use an image as source to create a Block Storage volume in the dashboard. This is the dashboard equivalent of the 'cinder upload-to-image' command.

This enhancement adds MAC address learning management to the Dashboard (horizon). Users are able to view and toggle the MAC address learning state of a port, in environments where this feature is supported.

The Block Storage scheduler now features a new volume number (count) weigher. This adds more flexibility in volume scheduling policy, as it allows you to configure volume distribution based on the number of volumes per back-end (rather than available back-end space).

The description for the 'Create Volume Type' dialogue has been enhanced to make it clear that creating a type is equivalent to the 'cinder type-create' command. After the volume type is created, the user can then further define the type by adding extra specs.

The "Format" field in the dashboard's Create Image window is now automatically populated after the user fills out the Image Source and Image File fields. Automatic population is based on the image file's extension.

This feature adds a configurable policy for managing SSH host keys, allowing system administrators to choose how secure they wish their SSH connections to be. As a result, the Block Storage service can store and verify SSH host keys, increasing security for drivers using SSH connections.

With this feature, the Compute service can now take snapshots of network-based Block Storage (cinder) volumes.

With this release, when retrieving a list of events from the Orchestration API, clients can now request a paginated list and avoid having to retrieve what may be a very large amount of data in a single request.

This feature improves the page loading performance when displaying the Project Volumes page.

This enhancement adds a SMBFS driver to Block Storage (cinder). As a result, Samba and Windows shares are now able to serve as volume backends.

With this enhancement, if OpenStack Networking is enabled, Packstack will display a warning if the Network Manager service is active on hosts.

With this release, the Orchestration API output when showing a stack now includes the username of the owner of a stack (i.e. the user who created it).

A transformer has been added that allows samples to be derived from primary measurements associated with different meters. Previously, derived samples could only be computed within a transformer based on the successive value of primary samples associated with the same meters. With this update, derived meters can now be constructed based on arithmetic rules involving multiple primary meters.

Support for OpenStack Networking DVR (Distributed Virtual Router) has been implemented in the dashboard (DVR is currently in technology preview).

OpenStack Networking DVR also includes new changes to the neutron CLI in the areas of router-creation, router-scheduling, and show commands (among others).

With this update, an administrator can now disable console access if the console is not accessible from outside the cloud infrastructure.

A configuration option has been added to the CONSOLE_TYPE parameter in the /etc/openstack_dashboard/local_settings file. Valid options are now: "AUTO", "VNC", "SPICE", "RDP", or "None". To disable console access, specify CONSOLE_TYPE="None".

A new volume driver has been added that supports SMB shares from Windows. The driver supports Windows Server 2012 (R2 included). Volumes are exported through SMB shares as disk images (similar flow to the NFS driver).

With this release, the Identity service now supports the standard JSON Home document format for API discovery. the service's previously supported API discovery response format was a non-standard JSON document. This format required a client to have Identity-specific knowledge to understand how to use the response. The JSON Home document format allows API discovery to use a standardized format, allowing client applications to perform API discovery across multiple services in a standard way.

The puppet-keystone puppet module now supports deploying the Identity service in Apache HTTPD.

With this update, a new feature has been added that enables to install OpenStack Identity service to run via Apache httpd processes. A new parameter 'CONFIG_KEYSTONE_SERVICE_NAME' has been added. Value 'httpd' will switch on Apache support while value 'keystone' allows Identity service run in it's own process as was implemented in the previous versions.

With this update, the dashboard supports Spark EDP jobs for OpenStack Data Processing (sahara); the user can now add "Spark" as a job type when creating jobs. To support the new job type, help text has been added for Spark-job creation, appropriate configuration fields are hidden when launching a Spark job, and the job-type list is now translatable.

Administrators can now create, update, or delete custom properties and metadata for images in the dashboard. This is useful for administrators and users to meaningfully describe images by sharing key-value pairs and tag metadata.

A new "Update metadata" image option in the Admin > Images panel now enables you to customize image properties and metadata.

When using nova-network with multiple networks, it is now possible to set the MTU, enable or disable DHCP, set the DHCP server, and indicate whether the network shares addresses with other networks. Previously, it was not possible to set these parameters on a per-network basis, making it more difficult to use nova-network with multiple networks. With this update, administrators now have more flexibility with settings when using multiple networks with nova-network.

In previous releases, the accuracy and timeliness of Telemetry samples could be negatively impacted if the central agent became overloaded by a large number of resources.

To mitigate this, the Telemetry service now features workload partitioning; this features allows the central agent to scale horizontally with each instance polling a disjointed set of resources. To do this, the 'tooz' utility coordinates group membership accross multiple central agents that share polling of resources.

With this update, operators can now disable L3 Router features using dashboard configuration. Two new settings have been added to the OPENSTACK_NEUTRON_NETWORK parameter in the /etc/openstack-dashboard/local_settings file:

* 'enable_router'. By default, "enable_router" is set to "True" because the router feature is enabled in most deployments. If "enable_router=False", the Router panel is hidden, and the 'Create Router' button as well as all routers are not displayed in the Network Topology map.

* 'enable_floatingip'. This setting handles situations where the L3 router extension is not present or router features should be hidden from users. If "enable_floatingip=False", the "Floating IP" tab and "Associate/Disassociate Floating IP" menu in the Instances table are hidden. This setting is only valid for deployments with using OpenStack Networking (neutron).

This release adds Block Storage Consistency Group support for the EMC VNX Direct Driver.

With this release, the "heat stack list" command now provides a "--show-nested" option that includes any nested stacks in the output. Normally, only stacks created directly by the user appear in the list and nested stacks are omitted.

Tooz-driven group membership coordination is now used, which allows multiple ceilometer-alarm-evaluator services to share the workload. The group membership-based solution provides a simple but robust technique for managing workload sharing that is less problematic than the previous RPC-fanout-based solution. Alarm evaluators can now be set up on multiple nodes using configuration for individual instances; if an evaluator fails, its workload is transferred to the other evaluators.

With this release, keystone now emits CADF notifications for role assignment events, providing a more complete audit trail. Role assignment operations affect a user's access to cloud resources; keeping an audit trail of these actions can be important to detect malicious actions.

The Orchestration service now features an OS::Heat::SoftwareDeployments resource, which can deploy a single SoftwareConfiguration to a group of servers. This is helpful for deploying a cluster where each server has to be configured with the IP addresses of every server in the cluster.

This feature updates cinderclient so that users are now prompted for a password via the command line if no password is specified via an environment variable or the --os-password option.

If no password is provided using either --os-password or env[OS_PASSWORD], python-glanceclient now falls back to password entry using tty.

To allow individual operations such as POST samples API to be selectively configured as admin-only, this update includes configurable Role-Based Access Control for the ceilometer API.

As a result, individual API operations can now be restricted to admin (or any other individual role) by adding a rule to the '/etc/ceilometer/policy.json' file of the form: "telemetry:create_samples": "rule:context_is_admin"

RBD snapshots and cloning are now used for Ceph-based ephemeral disk snapshots. With this update, data is manipulated within the Ceph server, rather than transferred across nodes, resulting in better snapshotting performance for Ceph.

This enhancement allows SR-IOV virtual functions (VF) to passthrough to 'flat' project network types. This is due to PCI passthrough with SR-IOV not being VLAN-specific. As a result, OpenStack Networking project networks with the "flat" network type can now take advantage of SR-IOV networking support.

Previously, the OS::Heat::ResourceGroup resource in a template failed validation if configured with size "0". Since some template authors use this mechanism to create optional parts of a template, a size of zero no longer causes a validation failure.

With this release, support for specifying a redis-sentinel in the backend URL had been added. This allows ceilometer fail over between a cluster of redis services, as to avoid the redis service being a single point of failure.

As a result, the tooz client can use a cluster of redis services as the backend with the sentinel failing over mastership of the cluster as necessary.

RBD snapshots and cloning are now used for Ceph-based ephemeral disk snapshots. With this update, data is manipulated within the Ceph server, rather than transferred across nodes, resulting in better snapshotting performance for Ceph.

OpenStack Data Processing (sahara) enables the fast provisioning and easy management of Hadoop clusters on OpenStack. Hadoop is used to store and analyze large amounts of data, which is usually unstructured but can be a combination of both complex and structured data.

OpenStack Data Processing is fully supported in this release. For information on how to install OpenStack Data Processing, refer to:

https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux_OpenStack_Platform/6/html/Installation_and_Configuration_Guide/chap-OpenStack_Sahara_Installation.html

The Block Storage service now allows you to set Consistency Groups. With this, you can group multiple volumes together as a single entity; this, in turn, allows you to perform operations on multiple volumes (for example, create snapshots) at once, rather than individually.

For this release, the Consistency Groups feature is only supported through the IBM Storwize driver.

This update introduces the rabbitmq-cluster resource agent for managing clustered RabbitMQ instances with the Pacemaker cluster manager.

This enhancement includes a feature, virt-v2v, which allows users the ability to convert images from a variety of hypervisors to run on OpenStack cloud.

This enhancement allows the passing of additional command-line options when creating an answer file. Previously, '--gen-answer-file' did not allow the specification of additional options. Instead, manual file editing was required to change any default options. With this update, it is now possible to combine '--gen-answer-file' with additional options, which are then included in the subsequently generated answer file.

This enhancement updates Packstack to retain temporary directories when running an installation in debug mode. This assists with troubleshooting activities, as retaining the temporary directories allows easier failure debugging. As a result, temporary directories are not deleted when running Packstack with the --debug command line option.

Capturing virtual machine memory usage is an useful metric to have in Ceilometer. This update includes a new enhancement that implements memory usage meter for virtual machines by inspecting libvirt 'virDomainMemoryStats'.

As a result, new usage metrics are captured by inspecting the relevant libvirt APIs.

In order to consume lifecycle events on network service resources such as firewall, load balancer and VPN services, this update adds a new enhancement for capturing network services lifecycle event notifications.

As a result, it includes new meters representing the support for consuming {FW|LB|VPN}aaS lifecycle events from OpenStack Networking.

To capture rate based disk metrics, this update adds transformation rules for per-device disk meters in order to derive per-second rates.

As a result, new derived rate based disk metrics and samples are generated.

This update includes the sahara-image-elements package, which contains tools for easily creating a Hortonworks Development Platform 2.0 image. This image is required for using the Data Processing (Sahara) feature in this release.

With this update, a new workaround (rhosp_disk_io_policies) configuration parameter is added to the 'nova.conf' file, allowing you to select the IO policy to use for various disk types. For example, you can now set rhosp_disk_io_policies=file=threads,network=native. This allows you to manually configure for performance.