Queens Series Release Notes

Queens Series Release Notes

17.0.5-48

Known Issues

  • The initial implementation of native LUKS decryption within Libvirt 2.2.0 had a known issue with the use of passphrases that were a multiple of 16 bytes in size. This was resolved in the upstream 3.3.0 release of Libvirt and has been backported to various downstream distribution specific versions.

    A simple warning will reference the above if this issue is encountered by Nova however operators of the environment will still need to update Libvirt to a version where this issue has been fixed to resolve the issue.

Upgrade Notes

  • A new online data migration has been added to populate missing instance.availability_zone values for instances older than Pike whose availability_zone was not specified during boot time. This can be run during the normal nova-manage db online_data_migrations routine. This fixes Bug 1768876

Security Issues

  • A new policy rule, os_compute_api:servers:create:zero_disk_flavor, has been introduced which defaults to rule:admin_or_owner for backward compatibility, but can be configured to make the compute API enforce that server create requests using a flavor with zero root disk must be volume-backed or fail with a 403 HTTPForbidden error.

    Allowing image-backed servers with a zero root disk flavor can be potentially hazardous if users are allowed to upload their own images, since an instance created with a zero root disk flavor gets its size from the image, which can be unexpectedly large and exhaust local disk on the compute host. See https://bugs.launchpad.net/nova/+bug/1739646 for more details.

    While this is introduced in a backward-compatible way, the default will be changed to rule:admin_api in a subsequent release. It is advised that you communicate this change to your users before turning on enforcement since it will result in a compute API behavior change.

  • To mitigate potential issues with compute nodes disabling themselves in response to failures that were either non-fatal or user-generated, the consecutive build failure counter functionality in the compute service has been changed to advise the scheduler of the count instead of self-disabling the service upon exceeding the threshold. The [compute]/consecutive_build_service_disable_threshold configuration option still controls whether the count is tracked, but the action taken on this value has been changed to a scheduler weigher. This allows the scheduler to be configured to weigh hosts with consecutive failures lower than other hosts, configured by the [filter_scheduler]/build_failure_weight_multiplier option. If the compute threshold option is nonzero, computes will report their failure count for the scheduler to consider. If the threshold value is zero, then computes will not report this value and the scheduler will assume the number of failures for non-reporting compute nodes to be zero. By default, the scheduler weigher is enabled and configured with a very large multiplier to ensure that hosts with consecutive failures are scored low by default.

Bug Fixes

  • Listing server and migration records used to give a 500 to users when a cell database was unreachable. Now only records from available cells are included to avoid the 500 error. The down cells are basically skipped when forming the results and this solution is planned to be further enhanced through the blueprint handling-down-cell.
  • For libvirt driver. Now when creating tap devices the MTU will be configured. Requires libvirt 3.3.0 at least.

17.0.5

Upgrade Notes

  • The nova-api service now requires the [placement] section to be configured in nova.conf if you are using a separate config file just for that service. This is because the nova-api service now needs to talk to the placement service in order to delete resource provider allocations when deleting an instance and the nova-compute service on which that instance is running is down. This change is idempotent if [placement] is not configured in nova-api but it will result in new warnings in the logs until configured. See bug https://bugs.launchpad.net/nova/+bug/1679750 for more details.
  • A new check is added to nova-status upgrade check which will scan all cells looking for nova-osapi_compute service versions which are from before Ocata and which may cause issues with how the compute API finds instances. This will result in a warning if:

    • No cell mappings are found
    • The minimum nova-osapi_compute service version is less than 15 in any given cell

    See https://bugs.launchpad.net/nova/+bug/1759316 for more details.

  • A new check is added to the nova-status upgrade check CLI which can assist with determining if ironic instances have had their embedded flavor migrated to use the corresponding ironic node custom resource class.

Security Issues

  • The ‘SSBD’ and ‘VIRT-SSBD’ cpu flags have been added to the list of available choices for the [libvirt]/cpu_model_extra_flags config option. These are important for proper mitigation of the Spectre 3a and 4 CVEs. Note that the use of either of these flags require updated packages below nova, including libvirt, qemu (specifically >=2.9.0 for virt-ssbd), linux, and system firmware. For more information see https://www.us-cert.gov/ncas/alerts/TA18-141A

Bug Fixes

  • The DELETE /os-services/{service_id} compute API will now return a 409 HTTPConflict response when trying to delete a nova-compute service which is still hosting instances. This is because doing so would orphan the compute node resource provider in the placement service on which those instances have resource allocations, which affects scheduling. See https://bugs.launchpad.net/nova/+bug/1763183 for more details.
  • The behaviour of ImagePropertiesFilter when using multiple architectures in a cloud can be unpredictable for a user if they forget to set the architecture property in their image. Nova now allows the deployer to specify a fallback in [filter_scheduler]image_properties_default_architecture to use a default architecture if none is specified. Without this, it is possible that a VM would get scheduled on a compute node that does not support the image.

17.0.4

Bug Fixes

  • The nova-compute service now allows specifying the interval for updating nova-compute-side cache of the compute node resource provider’s aggregates and traits info via a new config option called [compute]/resource_provider_association_refresh which defaults to 300. This was previously hard-coded to run every 300 seconds which may be too often in a large deployment.

17.0.3

Bug Fixes

  • The libvirt driver now allows specifying individual CPU feature flags for guests, via a new configuration attribute [libvirt]/cpu_model_extra_flags – only with custom as the [libvirt]/cpu_model. Refer to its documentation in nova.conf for usage details.

    One of the motivations for this is to alleviate the performance degradation (caused as a result of applying the “Meltdown” CVE fixes) for guests running with certain Intel-based virtual CPU models. This guest performance impact is reduced by exposing the CPU feature flag ‘PCID’ (“Process-Context ID”) to the guest CPU, assuming that it is available in the physical hardware itself.

    Note that besides custom, Nova’s libvirt driver has two other CPU modes: host-model (which is the default), and host-passthrough. Refer to the [libvirt]/cpu_model_extra_flags documentation for what to do when you are using either of those CPU modes in context of ‘PCID’.

17.0.2

Upgrade Notes

  • The image_ref_url entry in legacy instance notification payloads will be just the instance image id if [glance]/api_servers is not set and the notification is being sent from a periodic task. In this case the periodic task does not have a token to get the image service endpoint URL from the identity service so only the image id is in the payload. This does not affect versioned notifications.

Bug Fixes

  • The SchedulerReportClient (nova.scheduler.client.report.SchedulerReportClient) sends requests with the global request ID in the X-Openstack-Request-Id header to the placement service. Bug 1734625
  • The nova-manage discover_hosts command now has a --by-service option which allows discovering hosts in a cell purely by the presence of a nova-compute binary. At this point, there is no need to use this unless you’re using ironic, as it is less efficient. However, if you are using ironic, this allows discovery and mapping of hosts even when no ironic nodes are present.

17.0.0

Prelude

The 17.0.0 release includes many new features and bug fixes. It is difficult to cover all the changes that have been introduced. Please at least read the upgrade section which describes the required actions to upgrade your cloud from 16.0.0 (Pike) to 17.0.0 (Queens).

That said, a few major changes are worth mentioning. This is not an exhaustive list:

  • The latest Compute API microversion supported for Queens is v2.60. Details on REST API microversions added since the 16.0.0 Pike release can be found in the REST API Version History page.
  • The placement service should be upgraded before the nova controller and compute services. See the Pike Upgrade Notes for Queens for more details.
  • Some of the multi-cell cells v2 caveats have been resolved.
  • Cells v1 and nova-network continue to be deprecated, and plan to be removed in the 18.0.0 Rocky release.
  • The libvirt and xenapi compute drivers now have (experimental) native support for virtual GPU devices. See the virtual GPU admin guide for more details.
  • The libvirt compute driver now supports volume multi-attach when using the 2.60 compute API microversion. See the cinder admin guide for more details about volume multi-attach support in OpenStack.
  • Version 1.0.0 of the osc-placement plugin has been released which provides CLI support for interacting directly with the Placement API.
  • Traits-based scheduling is now available for the ironic compute driver. For more details, see the ironic docs for scheduling based on traits.

New Features

  • Added the StorPool libvirt volume attachment driver.
  • The libvirt driver now supports booting instances by asking for virtual GPUs. In order to support that, the operators should specify the enabled vGPU types in the nova-compute configuration file by using the configuration option [devices]/enabled_vgpu_types. Only the enabled vGPU types can be used by instances.

    For knowing which types the physical GPU driver supports for libvirt, the operator can look at the sysfs by doing:

    ls /sys/class/mdev_bus/<device>/mdev_supported_types
    

    Operators can specify a VGPU resource in a flavor by adding in the flavor’s extra specs:

    nova flavor-key <flavor-id> set resources:VGPU=1
    

    That said, Nova currently has some caveats for using vGPUs.

    • For the moment, only a single type can be supported across one compute node, which means that libvirt will create the vGPU by using that specific type only. It’s also possible to have two compute nodes having different types but there is no possibility yet to specify in the flavor which specific type we want to use for that instance.
    • Suspending a guest having vGPUs doesn’t work yet given a libvirt concern (it can’t hot-unplug mediated devices from a guest). Workarounds using other instance actions (like snapshotting the instance or shelving it) are recommended until libvirt supports that. If a user asks to suspend the instance, Nova will get an exception that will set the instance state back to ACTIVE, and you can see the suspend action in os-instance-action API will be Error.
    • Resizing an instance with a new flavor that has vGPU resources doesn’t allocate those vGPUs to the instance (the instance is created without vGPU resources). We propose to work around this problem by rebuilding the instance once it has been resized so then it will have allocated vGPUs.
    • Migrating an instance to another host will have the same problem as resize. In case you want to migrate an instance, make sure to rebuild it.
    • Rescuing an instance having vGPUs will mean that the rescue image won’t use the existing vGPUs. When unrescuing, it will use again the existing vGPUs that were allocated to the instance. That said, given Nova looks at all the allocated vGPUs when trying to find unallocated ones, there could be a race condition if an instance is rescued at the moment a new instance asking for vGPUs is created, because both instances could use the same vGPUs. If you want to rescue an instance, make sure to disable the host until we fix that in Nova.
    • Mediated devices that are created by the libvirt driver are not persisted upon reboot. Consequently, a guest startup would fail since the virtual device wouldn’t exist. In order to prevent that issue, when restarting the compute service, the libvirt driver now looks at all the guest XMLs to check if they have mediated devices, and if the mediated device no longer exists, then Nova recreates it by using the same UUID.
    • If you use NVIDIA GRID cards, please know that there is a limitation with the NVIDIA driver that prevents one guest to have more than one virtual GPU from the same physical card. One guest can have two or more virtual GPUs but then it requires each vGPU to be hosted by a separate physical card. Until that limitation is removed, please avoid creating flavors asking for more than one vGPU.

    We are working actively to remove or workaround those caveats, but please understand that for the moment this feature is experimental given all the above.

  • Add support, in new placement microversion 1.16, for a limit query parameter when making a GET /allocation_candidates request. The parameter accepts an integer value, N, which limits the number of candidates returned. A new configuration item [placement]/randomize_allocation_candidates, defaulting to False, controls how the limited results are chosen. If True, a random sampling of the entire result set is taken, otherwise the first N results are returned.
  • Add required query parameter to the GET /allocation_candidates API in new placement microversion 1.17. The parameter accepts a list of traits separated by ,, which is used to further limit the list of allocation requests to resource providers that have the capacity to fulfill the requested resources AND collectively have all of the required traits associated with them. In the same microversion, the provider summary includes the traits associated with each provider.
  • Add pagination support and changes-since filter for os-instance-actions API. Users can now use limit and marker to perform paginated query when listing instance actions. Users can also use changes-since filter to filter the results based on the last time the instance action was updated.
  • Added pagination support for migrations, there are four changes:
    • Add pagination support and changes-since filter for os-migrations API. Users can now use limit and marker to perform paginate query when listing migrations.
    • Users can also use changes-since filter to filter the results based on the last time the migration record was updated.
    • GET /os-migrations, GET /servers/{server_id}/migrations/{migration_id} and GET /servers/{server_id}/migrations will now return a uuid value in addition to the migrations id in the response.
    • The query parameter schema of the GET /os-migrations API no longer allows additional properties.
  • The 2.57 microversion makes the following changes:

    • The user_data parameter is added to the server rebuild API.
    • The personality parameter is removed from the server create and rebuild APIs. Use the user_data parameter instead.
    • The maxPersonality and maxPersonalitySize limits are excluded from the GET /limits API response.
    • The injected_files, injected_file_content_bytes and injected_file_path_bytes quotas are removed from the os-quota-sets and os-quota-class-sets APIs.

    See the spec for more details and reasoning.

  • When creating a baremetal instance with volumes, the ironic driver will now pass an IP address of an iSCSI initiator to the block storage service because the volume backend may require the IP address for access control. If an IP address is set to an ironic node as a volume connector resource, the address is used. If a node has MAC addresses as volume connector resources, an IP address is retrieved from VIFs associated with the MAC addresses. IPv4 addresses are given priority over IPv6 addresses if both are available.
  • A new param key_name was added to the instance rebuild API (v2.54), then it is able to reset instance key pair. It is worth noting that users within the same project are able to rebuild other user’s instances in that project with a new keypair. If set key_name to None in API body, nova will unset the keypair of instance during rebuild.
  • Added support for service create and destroy versioned notifications. The service.create notification will be emitted after the service is created (so the uuid is available) and also send the service.delete notification after the service is deleted.
  • The 1.12 version of the placement API changes handling of the PUT /allocations/{consumer_uuid} request to use a dict-based structure for the JSON request body to make it more aligned with the response body of GET /allocations/{consumer_uuid}. Because PUT requires user_id and project_id in the request body, these fields are added to the GET response. In addition, the response body for GET /allocation_candidates is updated so the allocations in the allocation_requests object work with the new PUT format.
  • Add support for graceful shutdown of VMware instances. The timeout parameter of the power_off method is now considered by the VMware driver. If you specify a timeout greater than 0, the driver calls the appropriate soft shutdown method of the VMware API and only forces a hard shutdown if the soft shutdown did not succeed before the timeout is reached.
  • The delete_host command has been added in nova-manage cell_v2 to delete a host from a cell (host mappings). The force option has been added in nova-manage cell_v2 delete_cell. If the force option is specified, a cell can be deleted even if the cell has hosts.
  • The nova-manage cell_v2 delete_cell command returns an exit code 4 when there are instance mappings to a cell to delete but all instances have been deleted in the cell.
  • Add a nova-manage cell_v2 list_hosts command for listing hosts in one or all v2 cells.
  • When cold migrating a server, the host parameter is available as of microversion 2.56. The target host is checked by the scheduler.
  • The following instance action records have been added:
    • attach_interface
    • detach_interface
    • attach_volume
    • detach_volume
    • swap_volume
    • lock
    • unlock
    • shelveOffload
    • createBackup
    • createImage
  • Microversion 2.55 adds a description field to the flavor resource in the following APIs:

    • GET /flavors
    • GET /flavors/detail
    • GET /flavors/{flavor_id}
    • POST /flavors
    • PUT /flavors/{flavor_id}

    The embedded flavor description will not be included in server representations.

    A new policy rule os_compute_api:os-flavor-manage:update is added to control access to the PUT /flavors/{flavor_id} API.

  • The payload of the instance.snapshot.start and instance.snapshot.end notifications have been extended with the snapshot_image_id field that contains the image id of the snapshot created. This change also causes that the type of the payload object has been changed from InstanceActionPayload version 1.5 to InstanceActionSnapshotPayload version 1.6. See the notification dev reference for the sample file of instance.snapshot.start as an example.
  • The following legacy notifications have been transformed to a new versioned payload:

    • aggregate.add_host
    • aggregate.remove_host
    • instance.evacuate
    • instance.interface_attach
    • instance.interface_detach
    • instance.live_migration_abort
    • instance.live_migration_pre
    • instance.rescue
    • instance.resize_confirm
    • instance.resize_prep
    • instance.resize_revert
    • instance.resize.error
    • instance.trigger_crash_dump
    • instance.unrescue
    • keypair.delete
    • keypair.import
    • server_group.create
    • server_group.delete

    Every versioned notification has a sample file stored under doc/notification_samples directory. Consult https://docs.openstack.org/nova/latest/reference/notifications.html for more information.

  • Configuration options for oslo.reports, found in the oslo_reports group, are now exposed in nova. These include:

    • log_dir
    • file_event_handler
    • file_event_handler_interval

    These will allow using a file trigger for the reports, which is particularly useful for Windows nodes where the default signals are not available. Also, specifying a log directory will allow the reports to be generated at a specific location instead of stdout.

  • A new 1.11 API microversion is added to the Placement REST API. This adds the resource_providers/{rp_uuid}/allocations link to the links section of the response from GET /resource_providers.
  • Throughout the Placement API, in microversion 1.15, ‘last-modified’ headers have been added to GET responses and those PUT and POST responses that have bodies. The value is either the actual last modified time of the most recently modified associated database entity or the current time if there is no direct mapping to the database. In addition, ‘cache-control: no-cache’ headers are added where the ‘last-modified’ header has been added to prevent inadvertent caching of resources.
  • New placement REST API microversion 1.14 is added to support nested resource providers. Users of the placement REST API can now pass a in_tree=<UUID> parameter to the GET /resource_providers REST API call. This will trigger the placement service to return all resource provider records within the “provider tree” of the resource provider with the supplied UUID value. The resource provider representation now includes a parent_provider_uuid value that indicates the UUID of the immediate parent resource provider, or null if the provider has no parent. For convenience, the resource provider resource also contains a root_provider_uuid field that is populated with the UUID of the top-most resource provider in the provider tree.
  • Microversion 1.13 of the Placement API gives the ability to set or clear allocations for more than one consumer uuid with a request to POST /allocations.
  • QEMU 2.6.0 and Libvirt 2.2.0 allow LUKS encrypted RAW files, block devices and network devices (such as rbd) to be decrypted natively by QEMU. If qemu >= 2.6.0 and libvirt >= 2.2.0 are installed and the volume encryption provider is ‘luks’, the libvirt driver will use native QEMU decryption for encrypted volumes. The libvirt driver will generate a secret to hold the LUKS passphrase for unlocking the volume and the volume driver will use the secret to generate the required encryption XML for the disk. QEMU will then be able to read from and write to the encrypted disk natively, without the need of os-brick encryptors.

    Instances that have attached encrypted volumes from before Queens will continue to use os-brick encryptors after a live migration or direct upgrade to Queens. A full reboot or another live migration between Queens compute hosts is required before the instance will attempt to use QEMU native LUKS decryption.

  • Now when you rebuild a baremetal instance, a new config drive will be generated for the node based on the passed in personality files, metadata, admin password, etc. This fix requires Ironic API 1.35.
  • Added traits support to the scheduler. A new flavor extra spec is added to support specifying the required traits. The syntax of extra spec is trait:<trait_name>=required, for example:

    • trait:HW_CPU_X86_AVX2=required
    • trait:STORAGE_DISK_SSD=required

    The scheduler will pass required traits to the GET /allocation_candidates endpoint in the Placement API to include only resource providers that can satisfy the required traits. Currently the only valid value is required. Any other value will be considered invalid.

    This requires that the Placement API version 1.17 is available before the nova-scheduler service can use this feature.

    The FilterScheduler is currently the only scheduler driver that supports this feature.

  • Added support for PCI device NUMA affinity policies. These allow you to configure how strict your NUMA affinity should be on a per-device basis or, more specifically, per device alias. These are configured as part of the [pci]alias configuration option(s).

    There are three policies supported:

    • required (must-have)
    • legacy (must-have-if-available) (default)
    • preferred (nice-to-have)

    In all cases, strict NUMA affinity is provided if possible. The key difference between the policies is how much NUMA affinity one is willing to forsake before failing to schedule.

  • This release adds support to attach a volume to multiple server instances. The feature can only be used in Nova if the volume is created in Cinder with the multiattach flag set to True. It is the responsibility of the user to use a proper filesystem in the guest that supports shared read/write access.

    This feature is currently only supported by the libvirt compute driver and only then if qemu<2.10 or libvirt>3.10 on the compute host.

    API restrictions:

    • Compute API microversion 2.60 must be used to create a server from a multiattach volume or to attach a multiattach volume to an existing server instance.
    • When creating a server using a multiattach volume, the API will check if the compute services have all been upgraded to a minimum level of support and will fail with a 409 HTTPConflict response if the computes are not yet upgraded.
    • Attaching a multiattach volume to a shelved offloaded instance is not supported and will result in a 400 HTTPBadRequest response.
    • Attaching a multiattach volume to an existing server instance will check that the compute hosting that instance is new enough to support it and has the capability to support it. If the compute cannot support the multiattach volume, a 409 HTTPConflict response is returned.

    See the Cinder enable multiattach spec for details on configuring Cinder for multiattach support.

  • Added a new boolean configuration option [filter_scheduler]shuffle_best_same_weighed_hosts (default is False).

    Enabling it will spread instances between hosts that have the same weight according to request spec. It is mostly useful when the [filter_scheduler]host_subset_size option has default value of 1, but available hosts have the same weight (e.g. ironic nodes using resource classes). In this case enabling it will decrease the number of rescheduling events.

    On the other hand, enabling it will make packing of VMs on hypervizors less dence even when host weighing is disabled.

  • When IP address substring filtering extension is available in Neutron, Nova will proxy the instance list with ip or ip6 filter to Neutron for better performance.
  • When using XenAPI driver for XenServer, we can support booting instances with a vGPU attached to get better graphics processing capability. In order to use this feature, the operators should specify the enabled vGPU types in the nova compute configuration file with the configuration option - [devices]/enabled_vgpu_types. Only the enabled vGPU types can be used by instances.

    XenServer automatically detects and groups together identical physical GPUs. Although the physical GPUs may support multiple vGPU types, at the moment nova only supports a single vGPU type for each compute node. The operators can run the following CLI commands in XenServer to get the available vGPU types if the host supports vGPU:

    xe vgpu-type-list
    

    The values of model-name ( RO): from the output of the above commands are the vGPU type names which you can choose from to set the nova configure - [devices]/enabled_vgpu_types. Please choose only one vGPU type to be enabled.

    The operators should specify a vGPU resource in the flavor’s extra_specs:

    nova flavor-key <flavor-id> set resources:VGPU=1
    

    Then users can use the flavor to boot instances with a vGPU attached. At the moment, XenServer doesn’t support multiple vGPUs for a single instance, so resources:VGPU in the flavor’s extra_specs should always be 1.

  • The VMware vCenter compute driver now supports booting from images which specify they require UEFI or BIOS firmware, using the hw_firmware_type image metadata.
  • VMware serial console log is completed. VSPC must be deployed along with nova-compute and configured properly. The [vmware]/serial_log_dir config option must have the same value in both nova.conf and vspc.conf.
  • Added a number of new configuration options to the [vnc] group, which together allow for the configuration of authentication used between the nova-novncproxy server and the compute node VNC server.

    • auth_schemes
    • vencrypt_client_key
    • vencrypt_client_cert
    • vencrypt_ca_certs

    For more information, refer to the documentation.

  • The nova-novncproxy server can now be configured to do a security negotiation with the compute node VNC server. If the VeNCrypt auth scheme is enabled, this establishes a TLS session to provide encryption of all data. The proxy will validate the x509 certs issued by the remote server to ensure it is connecting to a valid compute node. The proxy can also send its own x509 cert to allow the compute node to validate that the connection comes from the official proxy server.

    To make use of VeNCrypt, configuration steps are required for both the nova-novncproxy service and libvirt on all the compute nodes. The /etc/libvirt/qemu.conf file should be modified to set the vnc_tls option to 1, and optionally the vnc_tls_x509_verify option to 1. Certificates must also be deployed on the compute node.

    The nova.conf file should have the auth_schemes parameter in the vnc group set. If there are a mix of compute nodes, some with VeNCrypt enabled and others with it disabled, then the auth_schemes configuration option should be set to ['vencrypt', 'none'].

    Once all compute nodes have VeNCrypt enabled, the auth_schemes parameter can be set to just ['vencrypt'].

    For more information, refer to the documentation.

Known Issues

  • Due to a bug in python-glanceclient:

    https://bugs.launchpad.net/python-glanceclient/+bug/1707995

    If [glance]/api_servers is not set in nova.conf, and there is a versioned endpoint URL in the service catalog, nova makes a best attempt at parsing and stripping the version from the URL in order to make API requests to the image service.

  • Due to the changes in scheduling of bare metal nodes, additional resources may be reported as free to Placement. This happens in two cases:

    • An instance is deployed with a flavor smaller than a node (only possible when exact filters are not used)
    • Node properties were modified in ironic for a deployed nodes

    When such instances were deployed without using a custom resource class, it is possible for the scheduler to try deploying another instance on the same node. It will cause a failure in the compute and a scheduling retry.

    The recommended work around is to assign a resource class to all ironic nodes, and use it for scheduling of bare metal instances.

  • In 16.0.0 Pike release, quota limits are checked in a new fashion after change 5c90b25e49d47deb7dc6695333d9d5e46efe8665 and a new config option [quota]/recheck_quota has been added in change eab1d4b5cc6dd424c5c7dfd9989383a8e716cae5 to recheck quota after resource creation to prevent allowing quota to be exceeded as a result of racing requests. These changes could lead to requests blocked by over quota resulting in instances in the ERROR state, rather than no instance records as before. Refer to https://bugs.launchpad.net/nova/+bug/1716706 for detailed bug report.

Upgrade Notes

  • On AArch64 architecture cpu_mode for libvirt is set to host-passthrough by default.

    AArch64 currently lacks host-model support because neither libvirt nor QEMU are able to tell what the host CPU model exactly is and there is no CPU description code for ARM(64) at this point.

    Warning

    host-passthrough mode will completely break live migration, unless all the Compute nodes (running libvirtd) have identical CPUs.

  • A new keystone config section is added so that you can set session link attributes for communicating with keystone. This allows the use of custom certificates to secure the link between Nova and Keystone.
  • Starting in Ocata, there is a behavior change where aggregate-based overcommit ratios will no longer be honored during scheduling for the FilterScheduler. Instead, overcommit values must be set on a per-compute-node basis in the Nova configuration files.

    If you have been relying on per-aggregate overcommit, during your upgrade, you must change to using per-compute-node overcommit ratios in order for your scheduling behavior to stay consistent. Otherwise, you may notice increased NoValidHost scheduling failures as the aggregate-based overcommit is no longer being considered.

    You can safely remove the AggregateCoreFilter, AggregateRamFilter, and AggregateDiskFilter from your [filter_scheduler]enabled_filters and you do not need to replace them with any other core/ram/disk filters. The placement query in the FilterScheduler takes care of the core/ram/disk filtering, so CoreFilter, RamFilter, and DiskFilter are redundant.

    Please see the mailing list thread for more information: http://lists.openstack.org/pipermail/openstack-operators/2018-January/014748.html

  • The notify_on_api_faults config option and the api.fault notification it enabled have been removed. As noted in bug 1699115, the api.fault notification has not worked since the v2.1 API was introduced. As the v2.0 API is supported with the v2.1 codebase since Newton, this notification has not been emitted since Newton. Given that no one has reported an issue with this in that time, it is simply removed.
  • The nova-conductor service now needs access to the Placement service in the case of forcing a destination host during a live migration. Ensure the [placement] section of nova.conf for the nova-conductor service is filled in.
  • The [notifications]/default_publisher_id configuration option now defaults to [DEFAULT]/host rather than [DEFAULT]/my_ip.
  • Nova now requires Ceph/librados >= 11.1.0 if running under Python 3 with the RBD image backend for the libvirt driver. Requirements for Python 2 users or users using a different backend remain unchanged.
  • The default list of non-inherited image properties to pop when creating a snapshot has been extended to include image signature properties. The properties img_signature_hash_method, img_signature, img_signature_key_type and img_signature_certificate_uuid are no longer inherited by the snapshot image as they would otherwise result in a Glance attempting to verify the snapshot image with the signature of the original.
  • The TypeAffinityFilter, which was deprecated in the 16.0.0 Pike release, has been removed. The filter was flawed in that it relied on the flavor id primary key which cannot be relied upon since you cannot “edit” a flavor to change its disk, vcpu, etc values. Therefore to change a given flavor, it must be deleted and re-created, which means a new id value, thus potentially breaking the usefulness of the filter. Also, the flavor migration from the nova database to the nova_api database would also have resulted in different id values.
  • Nova no longer supports the Block Storage (Cinder) v2 API. Ensure the following configuration options are set properly for Cinder v3:
    • [cinder]/catalog_info - Already defaults to Cinder v3
    • [cinder]/endpoint_template - Not used by default.
  • If using the api_servers option in the [glance] configuration section, the values therein must be URLs. Raw IP addresses will result in hard failure of API requests.
  • Nova now uses keystoneauth1 configuration to set up communication with the image service. Use keystoneauth1 loading parameters for Session and Adapter setup in the [glance] conf section. This includes using endpoint_override in favor of api_servers. The [glance]api_servers conf option is still supported, but should only be used if you need multiple endpoints and are unable to use a load balancer for some reason. However, note that no configuration is necessary with an appropriate service catalog entry for the image service.
  • Nova now uses keystoneauth1 configuration to set up communication with the baremetal service. Use keystoneauth1 loading parameters for auth, Session, and Adapter setup in the [ironic] conf section. This includes using endpoint_override in favor of api_endpoint.
  • Nova now defaults to using the live snapshot feature of libvirt when taking snapshots. This was previously disabled by default due to crashes on load seen with older libvirt versions. It has been used in production by many clouds, and appears stable in real world environments. If you see crashes of guests during snapshots, you can disable this with the disable_libvirt_livesnapshot config value in [workarounds].
  • The minimum required version of shred has been increased to 8.22.
  • Nova now uses keystoneauth1 configuration to set up communication with the network service. Use keystoneauth1 loading parameters for Adapter setup in the [neutron] conf section (auth and Session options continue to work as before). Of note:
    • Legacy option url is deprecated and replaced by endpoint_override. This should not need to be specified if an appropriate service catalog entry exists for the network service.
    • When url is not used, region_name no longer defaults to RegionOne.
    • In Queens, specifying url will trigger the legacy behavior. The url option will be removed in Rocky.
  • Nova now uses keystoneauth1 configuration to set up communication with the placement service. Use keystoneauth1 loading parameters for auth, Session, and Adapter setup in the [placement] conf section. Note that, by default, the ‘internal’ interface will be tried first, followed by the ‘public’ interface. Use the conf option [placement].valid_interfaces to override this behavior.
  • A sys-admin privsep daemon has been added and needs to be included in your rootwrap configuration.
  • Calls to mount in the virt disk api no longer ignore the value of stderr.
  • The nova-idmapshift binary has been removed. This has been replaced by internal functionality using privsep.
  • The following commands are no longer required to be listed in your rootwrap configuration: blkid; blockdev; cat; chown; cryptsetup; dd; ebrctl; ifc_ctl; kpartx; losetup; lvcreate; lvremove; lvs; mkdir; mm-ctl; mount; nova-idmapshift; parted; ploop; prl_disk_tool; qemu-nbd; readlink; shred; tee; touch; umount; vgs; vrouter-port-control; and xend.
  • The following deprecated configuration options have been removed from the compute section of nova.conf:

    • null_kernel

    These were deprecated in the 15.0.0 release as they allowed for inconsistent API behavior across deployments.

  • The old deprecated keymgr options have been removed. Configuration options using the [keymgr] group will not be applied anymore. Use the [key_manager] group from Castellan instead. The Castellan api_class options should also be used instead, as most of the options that lived in Nova have migrated to Castellan.
    • Instead of api_class option nova.keymgr.barbican.BarbicanKeyManager, use castellan.key_manager.barbican_key_manager.BarbicanKeyManager
    • Instead of api_class option nova.tests.unit.keymgr.mock_key_mgr.MockKeyManager, use castellan.tests.unit.key_manager.mock_key_manager.MockKeyManager
    • nova.keymgr.conf_key_mgr.ConfKeyManager still remains, but the fixed_key configuration options should be moved to the [key_manager] section
  • Setting of ‘nicira-iface-id’ in XenServer VIF’s other-config field by XenAPI driver has been removed to avoid VM booting timeout problem when using neutron polling mode. It was previously deprecated in Pike.
  • The following nova-manage commands have been removed.

    • quota
    • shell
    • project
    • account
    • logs
    • host
    • agent

    These were previously deprecated in 16.0.0.

  • The following deprecated configuration options have been removed from the xenserver section of nova.conf:
    • remap_vbd_dev
    • remap_vbd_dev_prefix
  • The [DEFAULT] vendordata_driver option was deprecated in Mitaka and has now been removed. Configuration of vendordata drivers should now be done by using the [api] vendordata_providers option. For more information, refer to the vendordata documentation.
  • The [api] vendordata_providers option now defaults to [StaticJSON]. This ensures existing behavior of the vendordata v1 driver is preserved.
  • The TrustedFilter along with its related [trusted_computing] configuration options were deprecated in the 16.0.0 Pike release and have been removed in the 17.0.0 Queens release. The TrustedFilter was always experimental, had no continuous integration testing to prove it still worked, and no reported users.
  • The following configuration options have been renamed:

    • [vnc]vncserver_listen (now [vnc]server_listen)
    • [vnc]vncserver_proxyclient_address (now [vnc]server_proxyclient_address)

    This establishes a consistent naming between VNC and Spice options and removes some unnecessary duplication.

  • Nova now has a hard requirement that the Port Binding extension is enabled in the Neutron service. This simplifies the logic between Nova and Neutron.

Deprecation Notes

  • The idle_timeout option in the api_database group has been renamed to connection_recycle_time.
  • Nova API extension concept is removed in Pike. These extensions have their own policies enforcement which are not needed any more. All the below policies which were added mainly for extensions are deprecated for removal:
    • Show & List detail server
      • os_compute_api:os-config-drive
      • os_compute_api:os-extended-availability-zone
      • os_compute_api:os-extended-status
      • os_compute_api:os-extended-volumes
      • os_compute_api:os-keypairs
      • os_compute_api:os-server-usage
      • os_compute_api:os-security-groups (only from /servers APIs)
    • Create, Update, Show & List detail flavor
      • os_compute_api:os-flavor-rxtx
      • os_compute_api:os-flavor-access (only from /flavors APIs)
    • Show & List detail image
      • os_compute_api:image-size
  • The configuration options baremetal_enabled_filters and use_baremetal_filters are deprecated in Pike and should only be used if your deployment still contains nodes that have not had their resource_class attribute set. See Ironic release notes for upgrade concerns.
  • The following scheduler filters are deprecated in Pike: ExactRamFilter, ExactCoreFilter and ExactDiskFilter and should only be used if your deployment still contains nodes that have not had their resource_class attribute set. See Ironic release notes for upgrade concerns.
  • The hide_server_address_states configuration option is now deprecated for removal. In future, there will be hard coded server state building for which server address will be hidden. The policy ‘os_compute_api:os-hide-server-addresses’ also is deprecated for removal. More details here
  • The [glance]/allowed_direct_url_schemes configuration option and nova.image.download.modules extension point have been deprecated for removal. These were originally added for the nova.image.download.file FileTransfer extension which was removed in the 16.0.0 Pike release. The nova.image.download.modules extension point is not maintained and there is no indication of its use in production clouds. If you are using this extension point, please make the nova development team aware by contacting us in the #openstack-nova freenode IRC channel or on the openstack-dev mailing list.
  • The IronicHostManager is now deprecated along with the [scheduler]/host_manager option of ironic_host_manager.

    As of the 16.0.0 Pike release, the ExactRamFilter, ExactCoreFilter, and ExactDiskFilter scheduler filters are all deprecated along with the [scheduler]/use_baremetal_filters and [scheduler]/baremental_enabled_filters options. Deployments should migrate to using resource classes with baremetal flavors as described in the ironic install guide:

    https://docs.openstack.org/ironic/latest/install/configure-nova-flavors.html#scheduling-based-on-resource-classes

  • The following configuration options are deprecated for removal:

    • [DEFAULT]/monkey_patch
    • [DEFAULT]/monkey_patch_modules

    Monkey patching nova is not tested, not supported, and is a barrier to interoperability. If you have code which relies on monkey patching decorators, for example, for notifications, please propose those changes upstream.

    As a result, the following option is also deprecated for removal since it is only used when specified with [DEFAULT]/monkey_patch_modules:

    • [notifications]/default_publisher_id
  • The nova-manage cell command has been deprecated. This command configures various aspects of the Cells v1 functionality. Cells v1 has been deprecated, thus, this command is also deprecated. It will be removed in its entirety when Cells v1 is removed.
  • Scheduling bare metal (ironic) instances using standard resource classes (VCPU, memory, disk) is deprecated and will no longer be supported in Queens. Custom resource classes should be used instead. Please refer to the ironic documentation for a detailed explanation.
  • Support for Windows / Hyper-V Server 2012 has been deprecated in Queens in nova and will be removed in Rocky. The supported versions are Windows / Hyper-V Server 2012 R2 or newer.
  • Configuration option [ironic]api_endpoint is deprecated in favor of [ironic]endpoint_override.
  • Configuration options in the [placement] section are deprecated as follows:
    • os_region_name is deprecated in favor of region_name
    • os_interface is deprecated in favor of valid_interfaces

Security Issues

  • OSSA-2017-005: Nova Filter Scheduler bypass through rebuild action

    By rebuilding an instance, an authenticated user may be able to circumvent the FilterScheduler bypassing imposed filters (for example, the ImagePropertiesFilter or the IsolatedHostsFilter). All setups using the FilterScheduler (or CachingScheduler) are affected.

    The fix is in the nova-api and nova-conductor services.

  • OSSA-2017-006: Nova FilterScheduler doubles resource allocations during rebuild with new image (CVE-2017-17051)

    By repeatedly rebuilding an instance with new images, an authenticated user may consume untracked resources on a hypervisor host leading to a denial of service. This regression was introduced with the fix for OSSA-2017-005 (CVE-2017-16239), however, only Nova stable/pike or later deployments with that fix applied and relying on the default FilterScheduler are affected.

    The fix is in the nova-api and nova-scheduler services.

    Note

    The fix for errata in OSSA-2017-005 (CVE-2017-16239) will need to be applied in addition to this fix.

  • Privsep transitions. Nova is transitioning from using the older style rootwrap privilege escalation path to the new style Oslo privsep path. This should improve performance and security of Nova in the long term. - | privsep daemons are now started by nova when required. These daemons can be started via rootwrap if required. rootwrap configs therefore need to be updated to include new privsep daemon invocations.

Bug Fixes

  • A fix is made for bug 1482040 where a request to rebuild a volume-backed server with a new image which is different than what is in the root volume will now fail with a 400 Bad Request response. The compute API would previously return a 202 Accepted response but the backend compute service does not replace the image in the root disk so the API behavior was always wrong and is now explicit about the failure.
  • The fix for OSSA-2017-005 (CVE-2017-16239) was too far-reaching in that rebuilds can now fail based on scheduling filters that should not apply to rebuild. For example, a rebuild of an instance on a disabled compute host could fail whereas it would not before the fix for CVE-2017-16239. Similarly, rebuilding an instance on a host that is at capacity for vcpu, memory or disk could fail since the scheduler filters would treat it as a new build request even though the rebuild is not claiming new resources.

    Therefore this release contains a fix for those regressions in scheduling behavior on rebuild while maintaining the original fix for CVE-2017-16239.

    Note

    The fix relies on a RUN_ON_REBUILD variable which is checked for all scheduler filters during a rebuild. The reasoning behind the value for that variable depends on each filter. If you have out-of-tree scheduler filters, you will likely need to assess whether or not they need to override the default value (False) for the new variable.

  • Fixes bug 1695861 in which the aggregate API accepted requests that have availability zone names including ‘:’. With this fix, a creation of an availabilty zone whose name includes ‘:’ results in a 400 BadRequest error response.
  • When forcing a specified destination host during live migration, the scheduler is bypassed but resource allocations will still be made in the Placement service against the forced destination host. If the resource allocation against the destination host fails, the live migration operation will fail, regardless of the force flag being specified in the API. The guest will be unchanged on the source host. For more details, see bug 1712008.
  • When forcing a specified destination host during evacuate, the scheduler is bypassed but resource allocations will still be made in the Placement service against the forced destination host. If the resource allocation against the destination host fails, the evacuate operation will fail, regardless of the force flag being specified in the API. The guest will be unchanged on the source host. For more details, see bug 1713786.
  • The [DEFAULT]/log_options configuration option can be used to log configuration options at DEBUG level when the placement-api and/or nova-api services are started under WSGI. The default behavior is to log options on startup.
  • This release includes a fix for bug 1733886 which was a regression introduced in the 2.36 API microversion where the force parameter was missing from the PUT /os-quota-sets/{tenant_id} API request schema so users could not force quota updates with microversion 2.36 or later. The bug is now fixed so that the force parameter can once again be specified during quota updates. There is no new microversion for this change since it is an admin-only API.
  • If scheduling fails during rebuild the server instance will go to ERROR state and a fault will be recorded. Bug 1744325
  • It is now possible to configure the [cinder] section of nova.conf to allow setting admin-role credentials for scenarios where a user token is not available to perform actions on a volume. For example, when reclaim_instance_interval is a positive integer, instances are soft deleted until the nova-compute service periodic task removes them. If a soft deleted instance has volumes attached, the compute service needs to be able to detach and possibly delete the associated volumes, otherwise they will be orphaned in the block storage service. Similarly, if running_deleted_instance_poll_interval is set and running_deleted_instance_action = reap, then the compute service will need to be able to detach and possibly delete volumes attached to instances that are reaped. See bug 1733736 and bug 1734025 for more details.
  • The ironic virt driver no longer reports an empty inventory for bare metal nodes that have instances on them. Instead the custom resource class, VCPU, memory and disk are reported as they are configured on the node.
  • Fixes a bug preventing ironic nodes without VCPUs, memory or disk in their properties from being picked by nova.
  • The FilterScheduler now limits the number of results in the query it makes to placement to avoid situations where every compute node in a large deployment is returned. This is configurable with the new [scheduler]/max_placement_results configuration option, which defaults to 1000, a sane starting value for any size deployment.
  • It is now possible to unset the [vnc]keymap and [spice]keymap configuration options. These were known to cause issues for some users with non-US keyboards and may be deprecated in the future.
  • prevent swap_volume action if the instance is in state SUSPENDED, STOPPED or SOFT_DELETED. A conflict (409) will be raised now as previously it used to fail silently.
  • Fixes how memory stats are reported for VMware. The total memory for the vCenter cluster managed by Nova should be the aggregated sum of total memory of each ESX host in the cluster. This is more accurate than using the available memory of the resource pool associated to the cluster.
  • For the XenAPI driver, in order to delete cached images based on when they were created, a new --keep-days DAYS option is added to the destroy_cached_images script to delete cached images which were created at least DAYS days ago. By default, all unused cached images will be deleted when the script is run if they have cached_time.

Other Notes

  • Nova will now internally use a new flow for new volume attachments when:

    • All nova-compute services are upgraded
    • The block-storage 3.44 API microversion is available

    This change should be transparent to end users and does not affect existing volume attachments. Also, this does not affect how new volumes are created and attached during boot-from-volume when the block_device_mapping_v2.source_type is blank, image or snapshot and the block_device_mapping_v2.destination_type is volume.

    The motivation behind these changes are:

    • Track volume attachment state in the block storage service rather than the compute service (separation of duties, simplicity, etc)
    • Removal of technical debt from the compute service long-term
    • Enable a foundation on which to build support for multi-attach volumes

    More details can be found in the spec:

    https://specs.openstack.org/openstack/nova-specs/specs/queens/approved/cinder-new-attach-apis.html

  • Adds sata as a valid disk bus for qemu and kvm hypervisors. Setting the hw_disk_bus custom property on glance images allows for selecting the type of disk bus e.g. VIRTIO/IDE/SCSI. Some Linux (custom) images require use of SATA bus rather than any other that seem to be allowed.
  • The ironic driver will automatically migrate instance flavors for resource classes at runtime. If you are not able to run the compute and ironic services at pike because you are automating an upgrade past this release, you can use the nova-manage db ironic_flavor_migration to push the migration manually. This is only for advanced users taking on the risk of automating the process of upgrading through pike and is not recommended for normal users.
Creative Commons Attribution 3.0 License

Except where otherwise noted, this document is licensed under Creative Commons Attribution 3.0 License. See all OpenStack Legal Documents.