Rocky Series Release Notes

13.3.1-19

New Features

  • Adds log_filters parameter to nova::compute::libvirt. Defines a filter to select a different logging level for a given category log outputs, as specified in https://libvirt.org/logging.html . Default undef

  • Add tls_priority parameter to nova::compute::libvirt class to override the compile time default TLS priority string. The default is usually “NORMAL” unless overridden at build time. Only set this if it is desired for libvirt to deviate from the global default settings.

  • Add parameter mem_stats_period_seconds to specify number of seconds to memory usage statistics period, zero or negative value mean to disable memory usage statistics.

  • The libvirt driver allows passing different mount options to the nfs client. [libvirt]/nfs_mount_options, defaulting to <None>.

  • Add support for libvirt volume_use_multipath the ability to use multipath connection of the iSCSI or FC volume. Volumes can be connected in the LibVirt as multipath devices. Adds new parameter “nova::compute::libvirt::volume_use_multipath”

Upgrade Notes

  • Add support to enable/disable live-migration for NUMA topology instances

    Add configuration parameter workarounds/enable_numa_live_migration (defaults to false) which allows to enable/disable live-migration for NUMA topology instances.

Deprecation Notes

  • fping_path option is now deprecated for removal, the parameter has no effect.

Bug Fixes

  • Add cinder credentials in cinder section of nova conf to fix issue when reclaim_instance_interval > 0 and delete instance which booted from volume ‘delete_on_termination’ set as true. The volume status showing attached and in-use, even after instance deletion, since the admin context didn’t have any token info hence call to cinder api failed. In case when context is is_admin and without token,added cinder creds used to do authentication with user/project info to call cinder api. [Bug 1734025]

13.3.1

New Features

  • Added new parameter nova::scheduler::workers which can be used to set the amount of scheduler workers.

  • Adding nova::compute::sync_power_state_interval parameter that configure the interval to sync power states between the database and the hypervisor. Set to -1 to disable, or to 0 to run the default rate. Defaults to $::os_service_default.

Deprecation Notes

  • The image_service parameter is deprecated, as we are already using python-glanceclient instead of old glance client.

  • The notify_on_api_faults parameter is deprecated as it is also removed from nova config options in change https://review.openstack.org/#/c/505164.

13.3.0

New Features

  • Added parameters nova::ssl_only, nova::cert, nova::key to manage SSL options for VNC.

  • Add support for configuring ovsdb_connection config option, which can be used by ovs-vif repo in ovs-vsctl commands to reach remote ovsdb over tcp/ssl.

  • Add sync_power_state_pool_size as a parameter to nova::compute.

  • The nova_aggregate resource type can now unset aggregate metadata. [Bug 1776772]

  • Add support for configuring cpu_shared_set

    Some workloads run best when the hypervisor overhead processes (emulator threads in libvirt/QEMU) can be placed on different physical host CPUs than other guest CPU resources. This allow those workloads to prevent latency spikes for guest vCPU threads.

    To place a workload’s emulator threads on a set of isolated physical CPUs, set the configuration option to the set of host CPUs that should be used for best-effort CPU resources. Then set a flavor extra spec to hw:emulator_threads_policy=share to instruct nova to place that workload’s emulator threads on that set of host CPUs.

  • Adds nova::compute::libvirt_guests to manage /etc/sysconfig/libvirt-guests to configure libvirt-guests for compute reboot and handle running instance properly.

  • Class to serve Nova Metadata API with apache mod_wsgi in place of nova-metadata-api service. The new classes to configure are nova::metadata and nova::wsgi::apache_metadata.

  • Add parameters neutron_physnets_numa_nodes_mapping and neutron_tunnel_numa_nodes to provide numa affinity for physnets attached to vswitches.

  • Puppet-openstack now manages both [vnc]/enabled and [spice]/enabled. This was needed because Debian has [spice]/enabled set to True by default, and one cannot have both enabled in a compute node. Therefore, it was mandatory to have [spice]/enabled set to Flase if we’re using VNC.

  • Exposes the devices/enabled_vgpu_types config option for vgpu support. More information in the nova vgpu documentation.

Upgrade Notes

  • Users of puppet-openstack will have to set ::nova::compute::spice_enabled to True to make a meaningful decisioin to use Spice and not use VNC, and not rely on ::nova::spicehtml5proxy or ::nova::compute::spice to set it as enabled by default.

Deprecation Notes

  • The parameters nova::use_syslog, nova::use_stderr, nova::log_facility, nova::log_dir and nova::debug has been deprecated and moved to the nova::logging class, please use them there.

  • The following parameters gets deprecated in nova::api: nova_metadata_wsgi_enabled, neutron_metadata_proxy_shared_secret, metadata_cache_expiration, vendordata_jsonfile_path, vendordata_providers, vendordata_dynamic_targets, vendordata_dynamic_connect_timeout, vendordata_dynamic_read_timeout, vendordata_dynamic_failure_fatal, vendordata_dynamic_auth_auth_type, vendordata_dynamic_auth_auth_url, vendordata_dynamic_auth_os_region_name, vendordata_dynamic_auth_password, vendordata_dynamic_auth_project_domain_name, vendordata_dynamic_auth_project_name, vendordata_dynamic_auth_user_domain_name, vendordata_dynamic_auth_username

13.1.0

New Features

  • Add support for configuring cpu_weight_multiplier config option, which can be used to spread or stack workfloads on hosts based on their vCPU usage.

  • The libvirt driver now allows specifying individual CPU feature flags for guest, via a new configuration parameter [libvirt]/cpu_model_extra_flags – this is valid in combination with all the three possible values for [libvirt]/cpu_mode: custom, host-model and host-passthrough. The cpu_model_extra_flags allows specifying multiple CPU flags. The recent “Meltdown” CVE fixes have resulted in critical performance penalty which impacts every Nova guest within certain CPU models.

    More info is available in nova release notes.

  • Now puppet ensures that the nova-placement-api package is properly installed and that the service is running when running using Debian package.

  • Add support for configuring rx/tx queue sizes

    Adds support for configuring the rx_queue_size and tx_queue_size options in the QEMU virtio-net driver by way of nova.conf. Only supported for vhost/vhostuser interfaces. Currently, valid values for the ring buffer sizes are 256, 512, and 1024.

    Adjustable RX queue sizes requires QEMU 2.7.0, and libvirt 2.3.0 (or newer) Adjustable TX queue sizes requires QEMU 2.10.0, and libvirt 3.7.0 (or newer)

Upgrade Notes

  • The deprecated parameter nova::api::conductor_workers is removed. Please use nova::conductor::workers instead.

  • The deprecated parameter nova::api::osapi_max_limit is removed. Please use nova::api::max_limit instead.

  • The deprecated parameter nova::api::osapi_compute_link_prefix is removed. Please use nova::api::compute_link_prefix instead.

  • The deprecated parameter nova::api::osapi_glance_link_prefix is removed. Please use nova::api::glance_link_prefix instead.

  • The deprecated parameter nova::api::osapi_hide_server_address_states is removed. Please use nova::api::hide_server_address_states instead.

  • The deprecated parameter nova::api::default_floating_pool is removed. Please use nova::network::neutron::default_floating_pool instead.

  • The deprecated parameter nova::api::pci_alias is removed. Please use the nova::pci class instead and set the nova::pci::aliases parameter.

  • The deprecated parameter nova::migration::libvirt::use_tls is removed. Please set nova::migration::libvirt::transport to ‘tls’ instead if you had this option.

  • The deprecated class nova::cell_v2::cell is now removed. Please use the nova_cell_v2 resource type instead of using this class.

  • The deprecated nova::rabbit_host, nova::rabbit_hosts, nova::rabbit_password, nova::rabbit_port, nova::rabbit_userid and nova::rabbit_virtual_host is removed. Please use the nova::default_transport_url instead.

  • The deprecated nova::rpc_backend is removed.

  • The deprecated nova::metadata::novajoin::api::nova_user and nova::metadata::novajoin::api::nova_password is removed. Please use the nova::metadata::novajoin::api::service_user and nova::metadata::novajoin::api::service_password instead.

Deprecation Notes

  • nova::keystone::authtoken::auth_uri is deprecated and will be removed in a future release. Please use www_authenticate_uri instead.

  • nova::metadata::novajoin::authtoken::auth_uri is deprecated and will be removed in a future release. Please use www_authenticate_uri instead.

  • The nova scheduler baremetal Exact* filters and ironic host_manager have now been removed. As a result the following parameters are now deprecated and have no effect:

    nova::scheduler::filter::baremetal_scheduler_default_filter nova::scheduler::filter::scheduler_use_baremetal_filters nova::scheduler::filter::scheduler_host_manager

    Baremetal scheduling will use the custom resource class defined for each baremetal node to make its selection. Refer to the ironic documentation for more details:

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

Bug Fixes

  • Fix live_migration_inbound_addr when used with non-default port/user/extra_params

    live_migration_scheme was being used when live_migration_inbound_addr was set. This prohibits a non-default port/user/extra_params being used which is often required to fully describe the uri, particularly for the ssh transport (see https://bugs.launchpad.net/nova/+bug/1671288).

    However live_migration_inbound_addr also works with live_migration_uri, so for now revert back to using live_migration_uri in all cases.

13.0.0

New Features

  • Adds the pool_timeout option for configuring oslo.db. This will configure this value for pool_timeout with SQLAlchemy.

  • OpenStack Nova in Rocky has the ability of also purging the shadow data. This patch enables the cron configuration of this parameter for the archive rows Cron and adds an additional Cron to run the purge of the shadow tables without running the archive task.

  • Add the ability to set reserved_huge_pages on nova-computes via a string or a list of strings.

  • This patch will add a new parameter (age) to define a retention policy when purging the Nova shadow tables. Also will add the parameter all_cells (defaulted to false) to run the purge command over the cells tables.

  • Add openstack-db tag to Exec that run db-sync.

Deprecation Notes

  • nova::notify_api_faults is deprecated and will be removed in a future release. Please use nova::notify_on_api_faults instead.

  • os_region_name is deprecated and will be removed in a future release. Please region_name instead.

Bug Fixes

  • Fix the default values to saner ones, this is because operators might run both the archive and the purge cron jobs. The defaults will make the purge job to run each day at 5 hours but the retention policy parameter will retain the data for 14 days.