|
| 1 | +# eUPF Helm Chart |
| 2 | + |
| 3 | +Deploys [eUPF](https://github.com/edgecomllc/eupf) — an open-source 5G User Plane Function built on eBPF. |
| 4 | + |
| 5 | +## Quick start |
| 6 | + |
| 7 | +```bash |
| 8 | +helm install eupf .deploy/helm/eupf |
| 9 | +``` |
| 10 | + |
| 11 | +The default values produce a working pod: eUPF is configured via `UPF_*` environment variables (see below), probes and resource requests/limits are set, and the API is served on port 8080. |
| 12 | + |
| 13 | +## Cluster prerequisites |
| 14 | + |
| 15 | +- **Linux kernel >= 5.15** on the nodes (eBPF/XDP requirement). |
| 16 | +- **Unsafe sysctl allowed**: the chart sets `net.ipv4.ip_forward=1` in the pod security context, which nodes must allow: |
| 17 | + |
| 18 | + ``` |
| 19 | + kubelet --allowed-unsafe-sysctls 'net.ipv4.ip_forward' |
| 20 | + ``` |
| 21 | + |
| 22 | +- **Privileges**: eUPF loads eBPF programs, mounts bpffs/debugfs (via the image entrypoint) and raises `RLIMIT_MEMLOCK`. By default the chart runs the container as `privileged: true`. If your environment allows a finer-grained setup, override `securityContext`, e.g.: |
| 23 | + |
| 24 | + ```yaml |
| 25 | + securityContext: |
| 26 | + capabilities: |
| 27 | + add: [NET_ADMIN, SYS_ADMIN, SYS_RESOURCE, BPF, PERFMON] |
| 28 | + ``` |
| 29 | +
|
| 30 | +- The `/sys` hostPath is mounted read-only into the pod (required by the entrypoint to mount bpffs/debugfs). |
| 31 | + |
| 32 | +## Configuration |
| 33 | + |
| 34 | +eUPF supports configuration via CLI flags, environment variables and a YAML config file (see [docs/Configuration.md](https://github.com/edgecomllc/eupf/blob/main/docs/Configuration.md)). Precedence: CLI > **env** > config file > defaults. |
| 35 | + |
| 36 | +The chart supports two modes: |
| 37 | + |
| 38 | +### ENV mode (default, `config.enabled: false`) |
| 39 | + |
| 40 | +The chart injects a working default set of `UPF_*` variables (see `defaultEnv` in `values.yaml`), where `UPF_PFCP_NODE_ID` and `UPF_N3_ADDRESS` default to the pod IP. Override any of them via the `env` map: |
| 41 | + |
| 42 | +```yaml |
| 43 | +env: |
| 44 | + UPF_INTERFACE_NAME: eth0,n6 # comma-separated, no brackets |
| 45 | + UPF_LOGGING_LEVEL: debug |
| 46 | +``` |
| 47 | + |
| 48 | +List-valued parameters (e.g. `UPF_INTERFACE_NAME`) are comma-separated strings. |
| 49 | + |
| 50 | +### ConfigMap mode (`config.enabled: true`) |
| 51 | + |
| 52 | +```yaml |
| 53 | +config: |
| 54 | + enabled: true |
| 55 | + data: |
| 56 | + config.yml: | |
| 57 | + interface_name: eth0 |
| 58 | + api_address: :8080 |
| 59 | + pfcp_address: :8805 |
| 60 | + metrics_address: :9090 |
| 61 | +``` |
| 62 | + |
| 63 | +The chart creates a ConfigMap from `config.data`, mounts it at `/app/conf` and starts the container with `--config /app/conf/config.yml`. In this mode the default `UPF_*` env injection is **disabled**, because environment variables take precedence over the config file. You can still add env vars as targeted overrides via `env`. |
| 64 | + |
| 65 | +### Ports |
| 66 | + |
| 67 | +| Port | Name | Purpose | |
| 68 | +|------|------|---------| |
| 69 | +| 8080 | http | REST API (`service.port`, must match `UPF_API_ADDRESS` / `api_address`) | |
| 70 | +| 9090 | metrics | Prometheus metrics | |
| 71 | + |
| 72 | +Probes are TCP checks against the `http` port; keep `service.port` in sync with the API address if you change it. |
0 commit comments