Skip to content

Latest commit

 

History

History
113 lines (84 loc) · 5.92 KB

README.md

File metadata and controls

113 lines (84 loc) · 5.92 KB
Raw

Kubernetes on RISC-V

As of mid-2023, Debian and Ubuntu Linux run on RISC-V, with container support enabled in the kernel by default, however lack Kubernetes binaries. Previous work in the direction of running containers and Kubernetes on RISC-V indicates that it is possible. Back then, Golang support for RISC-V was at “experimental” status; now RISC-V is fully supported, so cross-compiling to the new architecture should be easier.

This document marks our current progress in making Kubernetes available for RISC-V.

Hello RISC-V!

Running

In summary, the following commands should get you up and running with K3s:

# Download
wget https://github.com/CARV-ICS-FORTH/k3s/releases/download/20241024/k3s-riscv64.gz.aa
wget https://github.com/CARV-ICS-FORTH/k3s/releases/download/20241024/k3s-riscv64.gz.ab
wget https://github.com/CARV-ICS-FORTH/k3s/releases/download/20241024/k3s-riscv64.gz.ac
cat k3s-riscv64.gz.* | gunzip > /usr/local/bin/k3s
chmod +x /usr/local/bin/k3s

# Install
curl -sfL https://get.k3s.io > k3s-install.sh
INSTALL_K3S_SKIP_DOWNLOAD="true" bash -x k3s-install.sh

Note

Most container images are not available for RISC-V, so many applications may not run.

Important

Currently, the bundled Traefik deployment requires running a version of iptables that uses the nftables API and migrating the CRDs to use the API Group traefik.io.

Check the examples folder for sample applications:

kubectl apply -f https://raw.githubusercontent.com/CARV-ICS-FORTH/kubernetes-riscv64/main/examples/hello-kubernetes.yaml

To install Argo Workflows:

kubectl create namespace argo
kubectl apply -n argo -f https://raw.githubusercontent.com/CARV-ICS-FORTH/kubernetes-riscv64/main/argo-workflows/install.yaml

K3s

K3s is a light-weight Kubernetes distribution that packs all necessary code into a single binary and needs a smaller memory footprint to run.

To cross-compile K3s to RISC-V, we also had to make required changes in its dependencies k3s-root (the base user space binaries for K3s) and runc (the tool that runs the containers).

Here is a list of submitted PRs:

Until RISC-V support is merged upstream in K3s, we maintain a fork with precompiled binaries. Each binary is built from the respective riscv64-release-* branch, that also uses available riscv64 container images by default.

We build K3s with:

# Build
rm -rf bin dist build
ARCH=riscv64 SKIP_IMAGE=true SKIP_VALIDATE=true SKIP_AIRGAP=true make

# Split
cd dist/artifacts
gzip < k3s-riscv64 | split -b 20M - k3s-riscv64.gz.

Supporting services and container images

K3s relies on several additional services and applications, which have been ported to RISC-V:

To build and upload the klipper-helm image from our fork, we run:

ARCH=riscv64 make
docker tag rancher/klipper-helm:dev carvicsforth/klipper-helm:v0.9.3-build20241008
docker push carvicsforth/klipper-helm:v0.9.3-build20241008

To build and upload the klipper-lb image from our fork, we run:

ARCH=riscv64 make
docker tag rancher/klipper-lb:dev carvicsforth/klipper-lb:v0.4.9
docker push carvicsforth/klipper-lb:v0.4.9

Other necessary container images, like the pause and metrics-server containers, can be found in the images folder.

Higher level services and applications

Higher level services ported to RISC-V:

To compile the Argo CLI for RISC-V, download our riscv64 branch of argo-workflows and run:

make dist/argo-linux-riscv64

To try it out:

cp /etc/rancher/k3s/k3s.yaml ~/.kube/config
argo submit -n argo --watch https://raw.githubusercontent.com/CARV-ICS-FORTH/kubernetes-riscv64/main/argo-workflows/hello-world.yaml

Other applications ported to RISC-V:

Acknowledgements

This project has received funding from the European Union’s Horizon Europe research and innovation programme through projects RISER ("RISC-V for Cloud Services", GA-101092993), AERO ("Accelerated EuRopean clOud", GA-101092850) and from the Key Digital Technologies Joint Undertaking through project REBECCA ("Reconfigurable Heterogeneous Highly Parallel Processing Platform for safe and secure AI", GA-101097224). KDT JU projects are jointly funded by the European Commission and the involved state members (including the Greek General Secretariat for Research and Innovation).