How to configure cluster nodes

Your cluster or multi-node appliance might contain nodes with different resources and different capacity. Therefore, each node can be configured separately.

Show node configuration

To display the current node configuration for a node (for example, lxd0), enter the following command:

amc node show <node>

This command will return output similar to the following:

name: lxd0
status: online
disk:
    size: 100GB
network:
    address: 10.119.216.34
    bridge-mtu: 1500
config:
    public-address: 10.119.216.34
    use-port-forwarding: true
    cpu-cores: 8
    cpu-allocation-rate: 4
    memory: 16GB
    memory-allocation-rate: 2
    gpu-slots: 10
    gpu-encoder-slots: 0
    tags: []

Configure allocation rates

AMS allows over-committing available resources on a node. This mechanism improves resource usage, because usually, containers don’t use 100% of their dedicated resources all of the time.

By default, AMS uses a CPU allocation rate of 4 and a memory allocation rate of 2. See Over-committing for more information.

You can configure the allocation rates with the cpu-allocation-rate and memory-allocation-rate configuration items.

Use the following commands to set the allocation rates on a node (for example, lxd0):

amc node set <node> cpu-allocation-rate <value>
amc node set <node> memory-allocation-rate <value>

Configure GPU slots

GPU slots are used to share GPUs amongst containers. See About GPU support and GPU slots for more information.

Each GPU-equipped cluster node is configured with a number of GPU slots and a number of GPU encoder slots. See Node-specific configuration for the default values that are used. Nodes without GPU are configured with 0 GPU slots and 0 GPU encoder slots.

Use the following command to change the number of GPU slots for a node (for example, lxd0):

amc node set <node> gpu-slots <number>

Last updated 5 months ago.