eva agent and related helm charts
eva-agent) and service account(eg. sa-eva-agent) for eva-agenteva-agent depends on eva-agent-ollama OR eva-agent-vllm. So they should be installed exclusively.
If you want to another release, don’t forget to fork another shell to avoid any confliction among env vars.
It includes values templates in addition to charts.
git clone https://github.com/mellerikat/eva-agent.git
cd eva-agent/release
CAUTION: Keep your local values not related to git repository. Refer .gitignore to make values directory name for your environment.
mkdir -p .values-${IMG_VER}
# eg. cp -drf 2.1-a1.1/1 .values-2.1-a1.1/1
cp -drf ${IMG_VER}/${REV} .values-${IMG_VER}/${REV}
Modify values in .values-${IMG_VER}/${REV}
cp .values-${IMG_VER}/${REV}/env.tpl .values-${IMG_VER}/${REV}/.env-${PLATFORM}
# Fill up empty env vars as following
# EA_PLATFORM=k3s
# EA_VALUE_ROOT=.values-2.1-a1.1/1
vi .values-${IMG_VER}/${REV}/.env-${PLATFORM}
source .values-${IMG_VER}/${REV}/.env-${PLATFORM}
source env_cli.sh
Now ea_* commands would run with values in .values-${IMG_VER}/${REV}
Initializes and defines resources for eva-agent package. Should be installed once.
ea_install eva-agent-init
ea_install eva-agent-qdrant
qdrant uses persistent storage, so PVC and PV remains after uninstall.
You should kubectl delete them manually if needed.
One inferrence engine should be installed to eva-agent image version.
ea_install eva-agent-vllm
vllm uses persistent storage, so PVC and PV remains after uninstall.
You should kubectl delete them manually if needed.
As you can see, nodeSelector should be defined for PVC initialization would be provisioned dynamically.
ea_install eva-agent-ollama
ollama uses persistent storage, so PVC and PV remains after uninstall.
You should kubectl delete them manually if needed.
In case of k3s, if you want to use persistence storage of host-path, you might apply overriding values as following.
ea_install eva-agent-ollama -f .values-${IMG_VER}/${REV}/eva-agent/values-k3s-bs.yaml
As you can see, nodeSelector should be defined for PVC initialization would be provisioned dynamically.
eva-agent is main service.
cp .values-${IMG_VER}/${REV}/eva-agent/secret-values.yaml.tpl \
.values-${IMG_VER}/${REV}/eva-agent/.secret-values.yaml
# Modify .secret-values.yaml
vi .values-${IMG_VER}/${REV}/eva-agent/.secret-values.yaml
ea_install eva-agent -f .values-${IMG_VER}/${REV}/eva-agent/.secret-values.yaml
# Requires AWS ECR pull permission.
# If you had not set 'aws configure', AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY would be asked.
# After enter them correctly, $HOME/.docker/config.json would be created or refreshed.
# $HOME/.docker/config.json would be transformed to $HOME/.docker/config-values.yaml and it would be used to imagePullSecret for eva-agent imamge.
ea_install is matched to helm install.
Likely, ea_template, ea_upgrade and ea_uninstall are available.
You might check actual helm command of COMMAND] as following.
$ ea_install eva-agent-init
...
=== install from repo chart
COMMAND] helm install eva-agent-init eva-agent/eva-agent-init --version=1.0.0 -n eva-agent -f .values-2.1-a1.1/1/eva-agent-init/values-k3s.yaml
NAME: eva-agent-init
LAST DEPLOYED: Fri Sep 26 12:43:28 2025
NAMESPACE: eva-agent
STATUS: deployed
REVISION: 1
TEST SUITE: None
ea_get_all is kubectl get all,pvc,pv -n $EA_NS.
# Install docker on k3s nodes if needed
# curl https://releases.rancher.com/install-docker/20.10.sh | sh
# Install k3s using docker not containerd
curl -sfL https://get.k3s.io | sudo sh -s - --docker
mkdir -p $HOME/.kube
sudo cp /etc/rancher/k3s/k3s.yaml $HOME/.kube/config
sudo chown $(id -un):$(id -gn) $HOME/.kube/config
# concat "export KUBECONFIG=$HOME/.kube/config" to $HOME/.bashrc and restart shell or "source $HOME/.bashrc"
kubectl version
# Client Version: v1.33.4+k3s1
# Kustomize Version: v5.6.0
# Server Version: v1.33.4+k3s1
sudo snap install helm --classic
Refer Appendix B. On-premise - nvidia gpu setup for k3s
Refer https://github.com/kubernetes-csi/csi-driver-nfs
helm repo add csi-driver-nfs https://raw.githubusercontent.com/kubernetes-csi/csi-driver-nfs/master/charts
helm repo update
helm install csi-driver-nfs csi-driver-nfs/csi-driver-nfs --namespace kube-system --version v4.11.0
sudo apt update
sudo apt install nfs-kernel-server -y
sudo mkdir -p /share/eva-agent
sudo chown nobody:nogroup /share/eva-agent
sudo chmod 777 /share/eva-agent
echo "/share/eva-agent *(rw,async,no_subtree_check,no_root_squash,insecure)" | sudo tee -a /etc/exports
sudo exportfs -ra
sudo systemctl restart nfs-kernel-server
# test
showmount -e localhost
sudo mkdir /mnt/tmp && sudo mount -t nfs -o rw,nfsvers=4 10.158.2.73:/share/eva-agent /mnt/tmp
sudo umount /mnt/tmp
kubectl create namespace eva-agent
kubectl create serviceaccount sa-eva-agent -n eva-agent
sudo /usr/local/bin/k3s-uninstall.sh
# manually or confirmed
sudo rm -rf /etc/rancher/k3s
sudo rm -rf /var/lib/rancher/k3s
sudo rm -rf /var/lib/kubelet
sudo rm -rf /etc/systemd/system/k3s*
# (Optional) Uninstall previous nvidia driver if needed
sudo apt update
sudo apt purge nvidia*
sudo apt update
sudo ubuntu-drivers autoinstall # recommend
sudo reboot
# Install nvidia container toolkit and cuda toolkit
# Refer:
# https://documentation.ubuntu.com/server/how-to/graphics/install-nvidia-drivers/index.html
# https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html
sudo apt-get update
sudo apt-get install -y nvidia-docker2
sudo nvidia-ctk runtime configure --runtime=docker
# Add "default-runtime": "nvidia" in /etc/docker/daemon.json as following
#
# {
# "default-runtime": "nvidia",
# "runtimes": {
# "nvidia": {
# "args": [],
# "path": "nvidia-container-runtime"
# }
# }
# }
sudo vi /etc/docker/daemon.json
sudo vi /var/lib/rancher/k3s/agent/etc/containerd/config.toml
# [plugins."io.containerd.grpc.v1.cri".containerd.runtimes.nvidia]
# runtime_type = "io.containerd.runc.v2"
# privileged_without_host_devices = false
# [plugins."io.containerd.grpc.v1.cri".containerd.runtimes.nvidia.options]
# BinaryName = "/usr/bin/nvidia-container-runtime"
sudo systemctl restart docker
sudo systemctl restart k3s # if needed
Check nvidia gpu setup as following and setup as your own way if needed.
which nvidia-ctk
# /usr/bin/nvidia-ctk
nvcc -V
# nvcc: NVIDIA (R) Cuda compiler driver
# Copyright (c) 2005-2021 NVIDIA Corporation
# Built on Thu_Nov_18_09:45:30_PST_2021
# Cuda compilation tools, release 11.5, V11.5.119
# Build cuda_11.5.r11.5/compiler.30672275_0
docker run --rm --gpus all nvidia/cuda:12.1.0-base-ubuntu22.04 nvidia-smi
# +-----------------------------------------------------------------------------------------+
# | NVIDIA-SMI 575.64.03 Driver Version: 575.64.03 CUDA Version: 12.9 |
# |-----------------------------------------+------------------------+----------------------+
# | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
# | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
# | | | MIG M. |
# |=========================================+========================+======================|
# | 0 NVIDIA GeForce GTX 1080 Off | 00000000:03:00.0 Off | N/A |
# | 28% 35C P8 6W / 180W | 40MiB / 8192MiB | 0% Default |
# | | | N/A |
# +-----------------------------------------+------------------------+----------------------+
# +-----------------------------------------------------------------------------------------+
# | Processes: |
# | GPU GI CI PID Type Process name GPU Memory |
# | ID ID Usage |
# |=========================================================================================|
# +-----------------------------------------------------------------------------------------+
# install
kubectl create -f https://raw.githubusercontent.com/NVIDIA/k8s-device-plugin/v0.17.1/deployments/static/nvidia-device-plugin.yml
# check
kubectl get all -n kube-system
# NAME READY STATUS RESTARTS AGE
# ...
# pod/nvidia-device-plugin-daemonset-9kp44 1/1 Running 0 37s
# ...
# NAME DESIRED CURRENT READY UP-TO-DATE AVAILABLE NODE SELECTOR AGE
# daemonset.apps/nvidia-device-plugin-daemonset 1 1 1 1 1 <none> 37s
kubectl logs pod/nvidia-device-plugin-daemonset-9kp44 -n kube-system
# ...
# I0910 06:21:55.093864 1 main.go:238] Starting FS watcher for /var/lib/kubelet/device-plugins
# I0910 06:21:55.093941 1 main.go:245] Starting OS watcher.
# I0910 06:21:55.094252 1 main.go:260] Starting Plugins.
# I0910 06:21:55.094285 1 main.go:317] Loading configuration.
# I0910 06:21:55.095184 1 main.go:342] Updating config with default resource matching patterns.
# I0910 06:21:55.095404 1 main.go:353]
# Running with config:
# ...
# I0910 06:21:55.095418 1 main.go:356] Retrieving plugins.
# I0910 06:21:55.121068 1 server.go:195] Starting GRPC server for 'nvidia.com/gpu'
# I0910 06:21:55.121858 1 server.go:139] Starting to serve 'nvidia.com/gpu' on /var/lib/kubelet/device-plugins/nvidia-gpu.sock
# I0910 06:21:55.124657 1 server.go:146] Registered device plugin for 'nvidia.com/gpu' with Kubelet
cat <<EOF | kubectl apply -f -
apiVersion: v1
kind: Pod
metadata:
name: gpu-pod
spec:
restartPolicy: Never
containers:
- name: cuda-container
image: nvcr.io/nvidia/k8s/cuda-sample:vectoradd-cuda12.5.0
resources:
limits:
nvidia.com/gpu: 1 # requesting 1 GPU
EOF
kubectl logs gpu-pod
# [Vector addition of 50000 elements]
# Copy input data from the host memory to the CUDA device
# CUDA kernel launch with 196 blocks of 256 threads
# Copy output data from the CUDA device to the host memory
# Test PASSED
# Done
kubectl delete pod gpu-pod
# uninstall
# kubectl delete -f https://raw.githubusercontent.com/NVIDIA/k8s-device-plugin/v0.17.1/deployments/static/nvidia-device-plugin.yml