On-premise Installation Guide

Requirements:

  • Kubernetes cluster:

    • It must be able to use OpenEBS Cstor. Disks have to be attached to Kubernetes nodes and must not be mounted or formatted.

    • If there is no internet access, each node should have a busybox:latest image preloaded.

  • A linux VM:

    • Must be accessible by the Kubernetes cluster (this VM will host the docker registry along with the chartmuseum and devpi services, which are needed to run the Neu.ro platform).

    • Must have access to the Kubernetes cluster.

    • The following utilities have to be installed: docker, kubectl, jq.

Archive Structure

/chartmuseum

  • A directory with all required helm charts. It will be mounted as a volume to the chartmuseum container.

/registry

  • A directory with all required docker images. It will be mounted as a volume to the registry container.

/devpi

  • A directory with the neuro-cli python package and all its dependencies. It will be mounted as a volume to the devpi container.

registry.tar

  • Saved registry:2 image.

chartmuseum.tar

  • Saved chartmuseum/chartmuseum:latest image.

devpi.tar

  • Saved devpi image.

jq.tar

  • Saved imega/jq:latest image, command-line JSON processor.

yq.tar

  • Saved mikefarah/yq:latest image, command-line YAML processor.

k8s/*.yaml

  • Kubernetes resources that will be created in the cluster.

*.sh

  • Installation scripts.

Platform Setup

Connect to the Linux VM and ensure that kubectl can connect to the Kubernetes cluster:

kubectl get nodes

Mount the USB (or external storage) device and extract the neuro.tar archive:

mkdir –p $HOME/neuro
tar -xvf neuro.tar -C $HOME/neuro

Prepare the config file (see example below), run the installation script, and wait until all pods are in the Running state:

$HOME/neuro/install.sh $CONFIG_FILE_PATH

By default, if there is no Ingress certificate specified in the config file, the installation script will generate a self-signed certificate. This self-signed certificate has to be added to the certificate trust store in the platform user's development environment.

Configure the DNS Server

Set up A records to the platform domains *.neu.ro, default.org.neu.ro, *.default.org.neu.ro, *.jobs.default.org.neu.ro in such a way that they point to all Kubernetes cluster IPv4 addresses.

Config File Example

server:
ip: "10.240.0.8"
ui:
type: minzdrav
ingress_ssl:
cert_path: "/path/to/ingress.crt" # optional
cert_key_path: "/path/to/ingress.key" # optional
postgres:
password: changeme
size: 10Gi
redis:
password: changeme
size: 10Gi
keycloak:
username: admin
password: changeme
auth:
jwt_secret: changeme
registry:
size: 10Gi
storage:
size: 10Gi
blob_storage:
size: 10Gi
metrics:
size: 10Gi
node_pools:
- name: cpu
cpu: 8
memory_gb: 6
disk_size_gb: 6
nodes:
- aks-agentpool-36699122-vmss000002
- name: gpu
cpu: 8
memory_gb: 6
disk_size_gb: 6
gpu: 1
gpu_model: nvidia-tesla-k80
nodes:
- aks-agentpool-36699122-vmss000002

Development Environment Setup

Add the certificate to the trust store (in case a self-signed certificate was generated during setup)

  • Download the Ingress certificate:

openssl s_client -connect app.neu.ro:443 -showcerts </dev/null > ingress.crt
  • Add it to your machine's trust store.

Install Neuro CLI

Run the following command to install Neuro CLI:

pip install -i http://$SERVER_IP/root/pypi neuro-cli