Edge configuration
An Edge Configuration tells Cast AI how to build the underlying compute instance for an edge node in a specific edge location. It covers settings that live below the Kubernetes layer — the boot disk, the machine image, the container runtime socket, and cloud instance tags — while node templates continue to control Kubernetes-level placement (constraints, labels, taints, and which edge locations are eligible).
This is conceptually the edge equivalent of a Node Configuration, scoped to a single edge location.
Before you begin
- OMNI must be enabled on your cluster. See Getting started with OMNI.
- You need at least one edge location already created. See Step 2: Create and onboard an edge location.
Edge Configurations and edge locations
Each Edge Configuration belongs to a single edge location. You can create more than one configuration per edge location — for example, one using a hardened custom image for production workloads and another using a stock distro image for testing — and mark one as the default, which is used whenever a specific configuration isn't otherwise requested.
| Concept | Scope | Configured by |
|---|---|---|
| Edge location | Cluster + region/provider | You, during onboarding |
| Edge Configuration | A single edge location | You, as described below |
| Node template | Cluster-wide, references edge locations | You, per Getting started with OMNI |
Create an Edge Configuration
- In the Cast AI console, open the edge location you want to configure and select Create edge node configuration.
- Fill in the fields in the drawer:
| Field | Description |
|---|---|
| Configuration name | A unique, descriptive name (for example, compute-optimized). This cannot be changed after saving, so name it deliberately. |
| Edge location | The edge location this configuration applies to. To configure additional edge locations, select them here and adjust their settings. |
| Instance tags | Key/value pairs applied to every Cast AI–provisioned instance built from this configuration. Use + Add to add more than one pair. These map to cloud-provider tags/labels (AWS tags, GCP labels, etc.) on the underlying instance. |
| Boot disk → Disk size | The boot disk size, in GiB, for instances built from this configuration. |
| Image | The machine image Cast AI should use when building the instance — for example, a GCP image path like projects/debian-cloud/global/images/debian-12-bookworm-v20240515, or the equivalent AWS AMI ID / OCI image OCID for other providers. Leave blank to use Cast AI's default image for the edge location's cloud provider. |
| CRI socket path | Path to the Container Runtime Interface socket on the node (for example, /run/containerd/containerd.sock). Only change this if your image uses a non-default container runtime or socket location. |
- Save the configuration.
NoteConfiguration names are permanent once saved. If you need to change other settings later, you can edit the existing configuration's other fields, or create a new configuration and switch your node template or manually created edge nodes to reference it instead.
Using an Edge Configuration
Once saved, an Edge Configuration is used automatically:
- If it's marked as the default for its edge location, it's used for any edge node Cast AI provisions there — whether through the Autoscaler or through manual edge provisioning — unless another configuration is explicitly requested.
- If it's not the default, it's used only when explicitly selected — for example, when manually provisioning an edge node and choosing a specific configuration.
Managing multiple configurations
To see all configurations for an edge location, open the edge location's detail page in the console. From there you can:
- Review each configuration's settings.
- Change which configuration is marked as default.
- Create additional configurations for different use cases (for example, GPU-optimized vs. general-purpose instances, or different images for production vs. testing).
Related documentation
Updated 17 hours ago