Openstack icehouse was officially announced in 4.17. After reading the release note, we found that the change was not small. It shows that openstack is still in rapid development, and many new features have been added and some are removed. The following is a simple translation of my projects, which has limited technical skills, limited horizons, and errors. I hope you can point it out.
1. Upgrade related:
1. Windows environment variables cannot be UNICODE, use PBR 0.8 to avoid this bug https://bugs.launchpad.net/pbr/+bug/1294246
2. Change log-config to log-config-append, and set disable_existing_loggers to false by default.
3. Add X-tenant-ID to prevent the user from manipulating the ports of other tenants and obtaining the host information of other tenants. Therefore, during the upgrade, you must first upgrade the network neutron, restart eutron-metadata-agent, and then upgrade the NOVA-compute. This reduces the time required.
Ii. openstack compute (NOVA)
1. Upgrade:
Currently, online upgrades are supported in certain programs. You can upgrade control nodes first, and then compute nodes one by one to reduce the service downtime caused by upgrades.
2. libvirt (KVM)
Libvirt (1) libvirt compute driver now supports modifying the number of kernel partitions to start the instance, which is obtained from the image's metadata OS _command_line field.
(2) libvirt driver now supports virtio SCSI (virtio-SCSI) instead of virtio block (virtio-BLK) to provide block storage devices. Virtio SCSI is a semi-virtualized SCSI controller designed for better scalability and performance.
(3) libvirt compute driver now supports adding a virtio RNG instance as the kernel entropy pool of the instance.
(4) libvirt driver now supports many other video drivers, such as VGA, cirrus, vmvga, xen, and qxl. Set the number of shards in the image metadate.
(5) Watchdog i6300esb is supported. The image's hw_watchdog_action attribute or flavor's extra_specs quota can be used to specify whether the instance is shut down, restarted, or sleep after a fault occurs.
(6) When you create an instance, disable hpet (High Precision Event timer), which is commonly known as a high-precision timer. Because this attribute is enabled, Windows will experience a clock offset during high load.
(7) you can wait for the neutron result when creating an instance. It seems that when the previous version number is used to create an instance, the instance status changes to error immediately because the network is not ready and 120 seconds have timed out.
3. API
(1) V3 APIs no longer support OS-DCF: diskconfig
(2) depracated is currently not recommended for XML format. The next version number will be officially abolished and can only be in JSON format.
(3) Use extended API: extendedservicesdelete to completely delete useless computing nodes;
(4) In V3, the Administrator permission operations are logically separated so that some of the functions can be used.
(5) The IP Address "| host_ip | 10.0.0.12" can be added to Nova hypervisor-show"
(6) Keystone V3 agrees that the tenant name is not unique. Therefore, when Nova interacts with neutron, the tenant uses the tenant ID instead of the tenant name to authenticate the identity.
4. Scheduler
(1) The caching scheduler driver is used to cache available host resources. Users have resource requests and directly use cached data for scheduling to improve the scheduling performance.
(2) added a new aggresponimagepropertiesisolation filter to filter hosts Based on the image attributes and aggregate (3) attributes. Two configuration items (/etc/NOVA. conf) are added ):
# Options defined in nova.scheduler.filters.aggregate_image_properties_isolation## Force the filter to consider only keys matching the given# namespace. (string value)#aggregate_image_properties_isolation_namespace=<None># The separator used between the namespace and keys (string# value)#aggregate_image_properties_isolation_separator=.
(4) When calculating weights, standardize the multiplier factor by 0.0 to 1.0.
(5) The scheduler supports the affinity and anti-affinity of instance groups. Instances are deployed according to the predefined policies.
5. Other features:
(1) There is a notification when keypairs is created or deleted.
(2) When the host status changes, a notification is sent, including enabled, disabled, powered on, shut down, and rebooted. The system exits the maintenance mode.
(3) When the compute service is stopped, new requests are no longer accepted, but the requests being processed are not affected, and the service is soft stopped.
(4) If the deleted instance is still being executed, add the number of partitions in the running_deleted_instance_action attribute: shutdown.
Old:
# Action to take if a running deleted instance is# detected.Valid options are ‘noop‘, ‘log‘ and ‘reap‘. Set to# ‘noop‘ to disable. (string value)#running_deleted_instance_action=log
New:
# Action to take if a running deleted instance is# detected.Valid options are ‘noop‘, ‘log‘, ‘shutdown‘, or# ‘reap‘. Set to ‘noop‘ to take no action. (string value)#running_deleted_instance_action=reap
(5) file injection is no longer supported by default. We recommend using configdrive or metadata. If you want to continue using the service, you must configure inject_key = true, inject_partition, and inject_password = true. These attributes are currently not recommended and may be unavailable in the future.
(6) Some configuration properties of/etc/NOVA. conf are renamed. You can see what the name is.
