# Device exposure

For [Jellyfin Hardware Acceleration](https://jellyfin.org/docs/general/administration/hardware-acceleration/), following instructions [here](https://github.com/kubernetes/kubernetes/issues/7890#issuecomment-766088805) (originally from [here](https://old.reddit.com/r/jellyfin/comments/i2r4h9/how_to_enable_hardware_acceleration_with_docker/)), I used [smarter-device-manager](https://gitlab.com/arm-research/smarter/smarter-device-manager) to expose devices from the host node (`epsilon`) into containers.

This was installed via a manual `kubectl apply`, though it should be migrated into GitOps-managed definitions - though I had to make some alterations to get ConfigMap to be read.

```yaml
# smarter-management-configmap.yaml
apiVersion: v1
kind: ConfigMap
metadata:
  name: smarter-device-manager
  namespace: smarter-device-management
data:
  conf.yaml: |
        - devicematch: ^fb0$
          nummaxdevices: 2

# smarter-management-pod.yaml
apiVersion: v1
kind: Pod
metadata:
  name: smarter-device-management
  namespace: smarter-device-management
spec:
  # Mark this pod as a critical add-on; when enabled, the critical add-on
  # scheduler reserves resources for critical add-on pods so that they can
  # be rescheduled after a failure.
  # See https://kubernetes.io/docs/tasks/administer-cluster/guaranteed-scheduling-critical-addon-pods/
  priorityClassName: "system-node-critical"
  hostNetwork: true
  dnsPolicy: ClusterFirstWithHostNet
  hostname: smarter-device-management
  nodeName: epsilon
  containers:
  - name: smarter-device-manager
    image: registry.gitlab.com/arm-research/smarter/smarter-device-manager:v1.20.11
    imagePullPolicy: IfNotPresent
    securityContext:
      allowPrivilegeEscalation: false
      capabilities:
        drop: ["ALL"]
    resources:
      limits:
        cpu: 100m
        memory: 10Mi
      requests:
        cpu: 10m
        memory: 10Mi
    volumeMounts:
      - name: device-plugin
        mountPath: /var/lib/kubelet/device-plugins
      - name: dev-dir
        mountPath: /dev
      - name: sys-dir
        mountPath: /sys
      - name: config
        mountPath: /root/config
  volumes:
    - name: device-plugin
      hostPath:
        path: /var/lib/kubelet/device-plugins
    - name: dev-dir
      hostPath:
            path: /dev
    - name: sys-dir
      hostPath:
            path: /sys
    - name: config
      configMap:
            name: smarter-device-manager
  terminationGracePeriodSeconds: 30
```

Re: `device-plugin` path, that apparently changed (from `/var/lib/rancher/k3s/agent/kubelet/device-plugins`, which was the provided value) [some time ago](https://github.com/k3s-io/k3s/issues/2664#issuecomment-742013918)

This also required the [Device Plugin Feature Gate](https://github.com/k3s-io/k3s/discussions/4596) to be enabled.

Further useful links:
* [Reddit thread](https://old.reddit.com/r/jellyfin/comments/y7i3uc/trouble_with_quicksync_trancoding_on_new_11th_gen/)
* [Enabling iGPU](https://community.hetzner.com/tutorials/howto-enable-igpu)

---

I spent a couple hours going down the rabbit-hole above, before noting that my server doesn't have an integrated graphics card, and so that was all for naught :) luckily, that is a problem that can be entirely solved with money (those are rare!) - a suitable card should arrive over the weekend and the hacking can continue.