Sean Mooney 1356ef5b57 Cyborg evacuate support

This change extends the conductor manager
to append the cyborg resource request to the
request spec when performing an evacuate.

This change passes the ARQs to spawn during rebuild
and evacuate. On evacuate the existing ARQs will be deleted
and new ARQs will be created and bound, during rebuild the
existing ARQs are reused.

This change extends the rebuild_instance compute rpcapi
function to carry the arq_uuids. This eliminates the
need to lookup the uuids associated with the arqs assinged
to the instance by quering cyborg.

Co-Authored-By: Wenping Song <songwenping@inspur.com>
Co-Authored-By: Brin Zhang <zhangbailin@inspur.com>

Implements: blueprint cyborg-rebuild-and-evacuate
Change-Id: I147bf4d95e6d86ff1f967a8ce37260730f21d236

2020-09-01 08:41:45 +00:00

2.6 KiB

Raw Blame History

Using accelerators with Cyborg

Starting from microversion 2.82, nova supports creating servers with accelerators provisioned with the Cyborg service, which provides lifecycle management for accelerators.

To launch servers with accelerators, the administrator (or an user with appropriate privileges) must do the following:

Create a device profile in Cyborg, which specifies what accelerator resources need to be provisioned. (See Cyborg device profiles API.)
Set the device profile name as an extra spec in a chosen flavor, with this syntax:
```
accel:device_profile=$device_profile_name
```
The chosen flavor may be a newly created one or an existing one.

Use that flavor to create a server:

openstack server create --flavor $myflavor --image $myimage $servername

Nova supports only specific operations for instances with accelerators. The lists of supported and unsupported operations are as below:

Supported operations.
- Creation and deletion.
- Reboots (soft and hard).
- Pause and unpause.
- Stop and start.
- Take a snapshot.
- Backup.
- Rescue and unrescue.
- Rebuild.
- Evacuate.
Unsupported operations
- Resize.
- Suspend and resume.
- Shelve and unshelve.
- Cold migration.
- Live migration.

22.0.0(Victoria)

Added support for rebuild and evacuate operations.

Some operations, such as lock and unlock, work as they are effectively no-ops for accelerators.

Caveats

Note

This information is correct as of the 21.0.0 Ussuri release. Where improvements have been made or issues fixed, they are noted per item.

For nested resource providers:

Creating servers with accelerators provisioned with the Cyborg service, if a flavor asks for resources that are provided by nested Resource Provider inventories (e.g. vGPU) and the user wants multi-create (i.e. say --max 2) then the scheduler could be returning a NoValidHosts exception even if each nested Resource Provider can support at least one specific instance, if the total wanted capacity is not supported by only one nested Resource Provider. (See bug 1874664.)

For example, creating servers with accelerators provisioned with the Cyborg service, if two children RPs have 4 vGPU inventories each:
- You can ask for a device profile in the flavor with 2 vGPU with --max 2.
- But you can't ask for a device profile in the flavor with 4 vGPU and --max 2.

2.6 KiB Raw Blame History

Using accelerators with Cyborg

Caveats

2.6 KiB

Raw Blame History