How to set up cluster groups¶

Cluster members can be assigned to Cluster groups. By default, all cluster members belong to the default group.

To create a cluster group, use the incus cluster group create command. For example:

incus cluster group create gpu

To assign a cluster member to one or more groups, use the incus cluster group assign command. This command removes the specified cluster member from all the cluster groups it currently is a member of and then adds it to the specified group or groups.

For example, to assign server1 to only the gpu group, use the following command:

incus cluster group assign server1 gpu

To assign server1 to the gpu group and also keep it in the default group, use the following command:

incus cluster group assign server1 default,gpu

To add a cluster member to a specific group without removing it from other groups, use the incus cluster group add command.

For example, to add server1 to the gpu group and also keep it in the default group, use the following command:

incus cluster group add server1 gpu

Configuration options¶

The following configuration options are available for cluster groups:

instances.vm.cpu.ARCHITECTURE.baseline

CPU base architecture name

Key:	`instances.vm.cpu.ARCHITECTURE.baseline`
Type:	string

The CPU base architecture name as can be found through qemu -cpu ?.

This can be a generic definition like qemu64 or kvm64, or it can be a specific hardware architecture like EPYC-v2. It’s important to ensure that all servers in the group match that baseline.

instances.vm.cpu.ARCHITECTURE.flags

CPU flags to add/remove to/from the baseline

Key:	`instances.vm.cpu.ARCHITECTURE.flags`
Type:	string

A comma separated list of CPU flags to add on top of CPU baseline or a list of flags to remove from it.

To remove a flag, use -flag.

user.*

Free form user key/value storage

Key:	`user.*`
Type:	string

User keys can be used in search.

Launch an instance on a cluster group member¶

With cluster groups, you can target an instance to run on one of the members of the cluster group, instead of targeting it to run on a specific member.

Note

scheduler.instance must be set to either all (the default) or group to allow instances to be targeted to a cluster group.

See Automatic placement of instances for more information.

To launch an instance on a member of a cluster group, follow the instructions in Launch an instance on a specific cluster member, but use the group name prefixed with @ for the --target flag. For example:

incus launch images:ubuntu/22.04 c1 --target=@gpu

Use with restricted projects¶

A project can be configured to only have access to servers that are part of specific cluster groups.

This is done by setting both restricted=true and restricted.cluster.groups to a comma separated list of group names.

Note

If the cluster group is renamed, the project restrictions will need to be updated for the new group name.