Containerize PacketRusher about packetrusher HOT 2 OPEN

I tried PacketRusher with a Dockerfile like this:

FROM golang:1.21-alpine3.18 AS build
RUN cd / \
 && wget -O PacketRusher.zip https://github.com/HewlettPackard/PacketRusher/archive/c10560874c0bbd8241f75c5aaa44536b528e1d74.zip \
 && unzip -n PacketRusher.zip \
 && mv PacketRusher-* PacketRusher \
 && cd /PacketRusher \
 && env CGO_ENABLED=0 go build cmd/packetrusher.go

FROM scratch
COPY --from=build /PacketRusher/packetrusher /
ENTRYPOINT ["/packetrusher"]

It works with a mounted configuration file, but there are several complications:

• Due to PacketRusher's use of vrf interface, the container must run in privileged mode.
  I'd recommend introducing a flag to disable vrf, so that the container can run with --cap-add=NET_ADMIN --device /dev/net/tun:/dev/net/tun.
  Applications who want to interact with the PDU session would need to explicitly bind to the netif that represents the PDU session.
• The multi-ue mode with --dedicatedGnbs flag expects the container to have several contiguous IPv4 addresses range.
  This may not be supported by Docker.
  I don't know of a clean solution to this.

from packetrusher.

Hi @yoursunny;

Thank you for your suggestion!

For 1., point taken, thanks for the suggestion. In that case, automatic routes creation should probably also be disabled when VRF are disabled. Else, funny things may happen on the host :-)

For 2., the use case we were thinking was k8s with Multus attaching multiple host interfaces (or creating multiple MacVLAN interfaces) and attaching it to the PacketRusher's pod. But I agree that having to use several contiguous IPv4 addresses is not the best in any cases. The switch to eUPF's eBPF from gtp5g as the WIP PR available here #47 should alleviate some of these issues but it's still WIP, and work only with a single UE for now. For PacketRusher on plain old Docker, I would recommend the usage of --net host, the usage of Docker's virtual interfaces may impact negatively the user plane performance of PacketRusher.

Thanks,
Valentin

from packetrusher.