# Token Factory Upgrade Instructions

Existing compute clusters created before `v3.1.39` must be updated using the latest bootstrap YAML after upgrading the controller to `v3.1.39`.

This update is required to enable:

- GPU type visibility
- GPU selection during model deployment
- Custom monitoring alerts

These steps apply only to existing compute clusters created before `v3.1.39`.

Newly created compute clusters do not require these steps.

* * *

## Step 1: Upgrade the Compute Cluster

Download the latest bootstrap YAML generated from the controller and re-apply it on the existing compute cluster.

```
kubectl apply -f <compute-name-compute>-bootstrap.yaml
```

During the upgrade process, services in the `gaap-controller` namespace are restarted, including:

- Endpoint pods
- Model deployment pods

Temporary service interruption may occur until all components become operational.

* * *

## Step 2: Scale Down Existing Model Deployments (If Required)

This step is required only when the compute cluster does not have sufficient resources to bring up new model deployment pods during rolling upgrades.

In such scenarios, the new model deployment pods may remain in the `Pending` state because the existing pods are already consuming the available resources.

Scale down the existing model deployment to `0` replicas.

```
kubectl scale deployment -n gaap-controller <model-deployment-name> --replicas=0
```

The `gaap-syncer` automatically reconciles and brings up the model deployments again.

* * *

## Post Upgrade Verification

After the upgrade is completed:

- GPU type information appears in the **GPU Inventory by Type** section
- GPU selection options are available during model deployment
- Custom monitoring alerts function as expected

* * *