Skip to Content
DocsQuick StartOpenShift

OpenShift

ecosystem.Ai can be installed on OpenShift. This can be tested locally using crc. Here we give example deployment configurations for the server, workbench, notebooks, runtime and grafana components.

Environment Variables

A number of environment variables can be used when starting up the ecosystem.Ai Deployments. These can be set in a ConfigMap. In addition, you should create a secret containing your license key.

✏️
Environment Variable Config Mapâ–·
apiVersion: v1
kind: ConfigMap
metadata:
  name: ecosystem-env
  namespace: ecosystem
data:
  # Server
  ECOSYSTEM_SERVER_PORT: "3001"
  ECOSYSTEM_SERVER_IP: "http://server.ecosystem.svc.cluster.local"
  ECOSYSTEM_PROP_FILE: "/config/ecosystem.properties"
  CLI_SETTINGS: "-Dserver.port=3001"
  RESET_USER: "true"
  NO_WORKBENCH: "true"
 
  # Runtime
  MONITORING_DELAY: "240"
  ECOSYSTEM_RUNTIME_PORT: "8091"
  NO_MONGODB: "true"
 
  # Workbench
  WORKBENCH_IP: /route for server deployment/
  WORKBENCH_PORT: "3001"
 
  # Grafana
  GF_SECURITY_ALLOW_EMBEDDING: "true"
  GF_INSTALL_PLUGINS: "marcusolsson-json-datasource,volkovlabs-echarts-panel"
✏️
Commands to Create Config Mapâ–·
oc apply -f ecosystem_configmap.yaml

Persistent Volume Claims

Ideally you should create a ReadWriteMany PVC to mount to the various ecosystem.Ai Deployments as it makes management easier. This is illustrated here. If your OpenShift instance does not support ReadWriteMany PVCs then you will need to create multiple PVCs.

✏️
Persistent Volume Claimâ–·
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: ecosystem-data-pvc
  namespace: ecosystem
spec:
  accessModes:
    - ReadWriteMany
  resources:
    requests:
      storage: /desired storage capacity/Gi
✏️
Commands to Create Persistent Volume Claimâ–·
oc apply -f ecosystem_pvc.yaml

Server Deployment

For the server, and in the subsequent components:

  • Images are pulled from Docker Hub. Generally, the images should be stored in a local repository and that repository should be referenced in the Deployment.
  • Resources are not specified as they will be specific to the environment.
✏️
Server Deploymentâ–·
apiVersion: apps/v1
kind: Deployment
metadata:
  name: ecosystem-server
  namespace: ecosystem
spec:
  replicas: 1
  selector:
    matchLabels:
      app: ecosystem-server
  template:
    metadata:
      labels:
        app: ecosystem-server
    spec:
      initContainers:
      - name: init-ecosystem-config
        image: registry.access.redhat.com/ubi9/ubi-minimal:latest
        command: ["/bin/sh", "-c"]
        args:
          - |
            cat << 'EOF' > /config/ecosystem.properties
            # ========
            logging.level=5
            cross.origin=ecosystem.ai
            date.format=yyyy-MM-dd'T'HH:mm:ss.SSSSSSXXX
            logger.level=500
            # these only needed for mongoexport and import
            mongo.ingestion.threads=7
            mongo.export.compress=true
            mongo.port=54445
            mongo.server=127.0.0.1
            mongo.data.port=54445
            mongo.data.server=127.0.0.1
            mongo.authentication.source=admin
            mongo.secure.data=true
            mongo.secure.workbench=true
            mongo.ecosystem.user=ecosystem_user
            mongo.ecosystem.password=EcoEco321
            mongo.profiles.user=ecosystem_user
            mongo.profiles.password=EcoEco321
            # this is primary connection string
            #user.profiles=profilesMaster
            user.profiles=ecosystem_meta
            mongo.connect=mongodb://ecosystem_user:EcoEco321@127.0.0.1:54445/?authSource=admin
            #mongo.connect=mongodb://ecosystem_user:EcoEco321@ecosystem-mongo:54445/?authSource=admin
            # prediction server
            prediction.server.h2o.mode=local
            prediction.server.h2o=http://127.0.0.1:54321
            prediction.server.pytorch={"train_model":"http://ecosystem-worker-pytorch:5010/train_model","training_status":"http://ecosystem-worker-pytorch:5010/get_model","download_model":"http://ecosystem-worker-pytorch:5010/download_model","get_model":"http://ecosystem-worker-pytorch:5010/get_model","load_model":"http://ecosystem-worker-pytorch:5000/load_model","get_current_model":"http://ecosystem-worker-pytorch:5000/get_current_model","score_document_info_formatted_h2o":"http://ecosystem-worker-pytorch:5000/score_document_info_formatted_h2o"}
            prediction.server.orbit={"server":"http://ecosystem-worker-orbit:5100/","connect":"mongodb://ecosystem_user:EcoEco321@ecosystem-server:54445/?authSource=admin"}
            prediction.server.prophet={"server":"http://ecosystem-worker-prophet:5110/","connect":"mongodb://ecosystem_user:EcoEco321@ecosystem-server:54445/?authSource=admin"}
            prediction.server.ecosystem=http://127.0.0.1:8080
            model.list.generative={"default":{"model":"llama3.2", "url":"http://host.docker.internal:11434/api/chat"}}
            # data and model storage
            user.home=./
            user.data=/data/
            user.generated.models=/data/models/
            user.deployed.models=/data/deployed/
            # apply for new application password in google: https://support.google.com/accounts/answer/185833
            user.email={"smtp":"smtp.gmail.com","port":587,"login":"","password":"","admin":"","cc":"","rule":"super_only"}
            # used for code generation of runtime
            sourcecontrol.runtime=[{"server":"https://github.com/ecogenetic/ecosystem-runtime.git","branch":"workbench","user":"ecosystemai","password":""}]
            # presto server
            presto.url=jdbc:trino://ecosystem-worker-trino:8084/
            presto.connection=local/master?user=admin
            logging.database=logging
            logging.collection=ecosystemruntime
            logging.collection.response=ecosystemruntime_response
            EOF
        volumeMounts:
          - name: ecosystem-data
            subPath: serverconfig
            mountPath: /config
      containers:
      - name: ecosystem-server
        image: docker.io/ecosystemai/ecosystem-server:arm64
        imagePullPolicy: IfNotPresent
        env:
          - name: MASTER_KEY
            valueFrom:
              secretKeyRef:
                name: master-key
                key: master-key
          - name: PORT
            valueFrom:
              configMapKeyRef:
                name: ecosystem-env
                key: ECOSYSTEM_SERVER_PORT
          - name: IP
            valueFrom:
              configMapKeyRef:
                name: ecosystem-env
                key: ECOSYSTEM_SERVER_IP
          - name: ECOSYSTEM_PROP_FILE
            valueFrom:
              configMapKeyRef:
                name: ecosystem-env
                key: ECOSYSTEM_PROP_FILE
          - name: CLI_SETTINGS
            valueFrom:
              configMapKeyRef:
                name: ecosystem-env
                key: CLI_SETTINGS
          - name: RESET_USER
            valueFrom:
              configMapKeyRef:
                name: ecosystem-env
                key: RESET_USER
          - name: NO_WORKBENCH
            valueFrom:
              configMapKeyRef:
                name: ecosystem-env
                key: NO_WORKBENCH
        ports:
          - containerPort: 3001
            name: http
          - containerPort: 54321
            name: htwoo
          - containerPort: 54445
            name: mongo
        volumeMounts:
          - name: ecosystem-data
            subPath: data
            mountPath: /data
          - name: ecosystem-data
            subPath: serverconfig
            mountPath: /config
      volumes:
      - name: ecosystem-data
        persistentVolumeClaim:
          claimName: ecosystem-data-pvc
✏️
Commands to Create Serverâ–·
oc apply -f server-deployment-openshift.yaml
 
oc expose deployment ecosystem-server --port=3001 --name=server
oc expose deployment ecosystem-server --port=54445 --name=mongo
oc expose deployment ecosystem-server --port=54321 --name=htwoo
oc expose svc server --port=3001
oc expose svc mongo --port=54445
oc expose svc htwoo --port=54321

Notebooks and Grafana Deployments

For the notebooks deployment, we configure runAsUser: 1001570001. This is a UID with a pre-created user in the Jupyter Notebooks environment. We pre-create the user as user creation will require the pod to run as root. If this runAsUser is not allowed by your serviceAccount then you will need to contact ecosystem.Ai to have an allowed UID preconfigured as a user.

The grafana deployment uses the nojwt image tag. This requires some manual configuration of the connection between grafana and the server. Automating the connection between grafana and the server can cause permission issues in OpenShift.

✏️
Notebooks and Grafana Deploymentsâ–·
apiVersion: apps/v1
kind: Deployment
metadata:
  name: ecosystem-notebooks
  namespace: ecosystem
spec:
  replicas: 1
  selector:
    matchLabels:
      app: ecosystem-notebooks
  template:
    metadata:
      labels:
        app: ecosystem-notebooks
    spec:
      containers:
      - name: ecosystem-notebooks
        securityContext:
          runAsUser: 1001570001
        image: docker.io/ecosystemai/ecosystem-notebooks:latest
        imagePullPolicy: IfNotPresent
        ports:
          - containerPort: 8000
            name: notebooks
          - containerPort: 8010
            name: alt
        volumeMounts:
          - name: ecosystem-data
            mountPath: "/app/Shared Projects"
            subPath: "notebooks-users/notebooks"
          - name: ecosystem-data
            mountPath: "/home"
            subPath: "notebooks-users"
          - name: ecosystem-data
            subPath: data
            mountPath: "/data"
      volumes:
        - name: ecosystem-data
          persistentVolumeClaim:
            claimName: ecosystem-data-pvc
      restartPolicy: Always
 
---
 
apiVersion: apps/v1
kind: Deployment
metadata:
  name: ecosystem-grafana
  namespace: ecosystem
spec:
  replicas: 1
  selector:
    matchLabels:
      app: ecosystem-grafana
  template:
    metadata:
      labels:
        app: ecosystem-grafana
    spec:
      containers:
      - name: ecosystem-grafana
        image: docker.io/ecosystemai/ecosystem-grafana:nojwt
        imagePullPolicy: IfNotPresent
        env:
          - name: GF_SECURITY_ALLOW_EMBEDDING
            valueFrom:
              configMapKeyRef:
                name: ecosystem-env
                key: GF_SECURITY_ALLOW_EMBEDDING
          - name: $GF_AUTH_JWT_URL
            valueFrom:
              configMapKeyRef:
                name: ecosystem-env
                key: GF_AUTH_JWT_URL
          - name: $GF_AUTH_USERNAME
            valueFrom:
              configMapKeyRef:
                name: ecosystem-env
                key: GF_AUTH_USERNAME
          - name: $GF_AUTH_PASSWORD
            valueFrom:
              configMapKeyRef:
                name: ecosystem-env
                key: GF_AUTH_PASSWORD
        ports:
          - containerPort: 3000
            name: http
        volumeMounts:
          - name: ecosystem-data
            subPath: grafana
            mountPath: /var/lib/grafana
      volumes:
        - name: ecosystem-data
          persistentVolumeClaim:
            claimName: ecosystem-data-pvc
      restartPolicy: Always
✏️
Commands to Create Notebooks and Grafanaâ–·
oc apply -f notebooks-grafana-deployment-openshift.yaml
 
oc expose deployment ecosystem-grafana --port=3000
oc expose svc ecosystem-grafana --port=3000
 
oc expose deployment ecosystem-notebooks --port=8000 --name=jupyter
oc expose deployment ecosystem-notebooks --port=8010 --name=pythonserver
oc expose svc jupyter --port=8000
oc expose svc pythonserver --port=8010

Workbench and Runtime Deployments

The workbench Deployment utilises a ConfigMap which sets the port on which the workbench starts up. In this case we start up the workbench on port 8008.

A single runtime Deployment is created. Separate Deployments should be created for each use case that needs to be pushed to a runtime. The replica count is not specified in the runtime Deployment, this assumes that a HorizontalPodAutoscaler will be created for the runtime. Alternatively the replica count can be set at a volume that can handle the anticipated load.

✏️
Workbench and Runtime Deploymentsâ–·
apiVersion: v1
kind: ConfigMap
metadata:
  name: nginx-config
  namespace: ecosystem
data:  
  default.conf: |
    server {
        listen 8008 default_server;
        listen [::]:8008 default_server;
 
        root /usr/share/nginx/html;
 
        index index.html;
 
        server_name _;
 
        location / {
            try_files $uri$args $uri$args/ /index.html;
        }
 
        location ~* .(js|css|ttf|ttc|otf|eot|woff|woff2)$ {
                add_header access-control-allow-origin "*";
                expires max;
        }
    }
 
---
 
apiVersion: apps/v1
kind: Deployment
metadata:
  name: ecosystem-workbench
  namespace: ecosystem
spec:
  replicas: 1
  selector:
    matchLabels:
      app: ecosystem-workbench
  template:
    metadata:
      labels:
        app: ecosystem-workbench
    spec:
      containers:
      - name: ecosystem-workbench
        image: docker.io/ecosystemai/ecosystem-workbench:latest
        volumeMounts:
        - name: nginx-config
          mountPath: "/etc/nginx/conf.d"
        securityContext:
          runAsGroup: 0
        imagePullPolicy: IfNotPresent
        env:
          - name: IP
            valueFrom:
              configMapKeyRef:
                name: ecosystem-env
                key: WORKBENCH_IP
          - name: PORT
            valueFrom:
              configMapKeyRef:
                name: ecosystem-env
                key: WORKBENCH_PORT
        ports:
          - containerPort: 8008
            name: http
      volumes:
      - name: nginx-config
        configMap:
          name: nginx-config
      restartPolicy: Always
 
---
 
apiVersion: apps/v1
kind: Deployment
metadata:
  name: ecosystem-runtime1
  namespace: ecosystem
spec:
  replicas: 1
  selector:
    matchLabels:
      app: ecosystem-runtime1
  template:
    metadata:
      labels:
        app: ecosystem-runtime1
    spec:
      containers:
      - name: ecosystem-runtime1
        image: docker.io/ecosystemai/ecosystem-runtime-solo:latest
        imagePullPolicy: IfNotPresent
        volumeMounts:
          - name: ecosystem-data
            subPath: data
            mountPath: /data
        env:
          - name: MASTER_KEY
            valueFrom:
              secretKeyRef:
                name: master-key
                key: master-key
          - name: NO_MONGODB
            valueFrom:
              configMapKeyRef:
                name: ecosystem-env
                key: NO_MONGODB
          - name: MONITORING_DELAY
            valueFrom:
              configMapKeyRef:
                name: ecosystem-env
                key: MONITORING_DELAY
          - name: PORT
            valueFrom:
              configMapKeyRef:
                name: ecosystem-env
                key: PORT
        ports:
          - containerPort: 8091
            name: http
      volumes:
      - name: ecosystem-data
        persistentVolumeClaim:
          claimName: ecosystem-data-pvc
      restartPolicy: Always
✏️
Commands to Create Workbench and Runtimeâ–·
oc apply -f workbench-runtime-deployment-openshift.yaml
 
oc expose deployment ecosystem-workbench --port=8008
oc expose svc ecosystem-workbench --port=8008
 
oc expose deployment ecosystem-runtime1 --port=8091
oc expose svc ecosystem-runtime1 --port=8091

Workbench2 Deployments

The workbench2 incorporates the generative AI enabled UI and the notebooks functionality through which the python package can be used.

✏️
Workbench2 Deploymentsâ–·
apiVersion: v1
kind: ConfigMap
metadata:
  name: workbench2-config
  namespace: ecosystem
data:
  MONGODB_URI: "mongo_connection_string"
  MONGODB_DATABASE: "ecosystem_meta"
  ECOSYSTEM_SERVER: "http://server-ecosystem.apps-crc.testing"
  BACKEND_PORT: "8001"
  CORS_ORIGINS: "http://localhost:5270,http://workbench2.ecosystem.svc.cluster.local:5270"
  LOG_LEVEL: "INFO"
  H2O_URL: "http://htwoo.ecosystem.svc.cluster.local:54321"
  H2O_CONTAINER_DATA_PATH: "/data"
  LOCAL_DATA_PATH: "/data"
  H2O_MODELS: "/data/models"
  H2O_SCORER_PORT: "9090"
  H2O_MODELS_DIR: "/data/models"
  H2O_OFFLINE: "http://localhost:9090"
  H2O_PREFER_OFFLINE_SCORING: "true"
  TRINO_HOST: "trino.ecosystem.svc.cluster.local"
  TRINO_PORT: "8080"
  TRINO_USER: "ecosystem_user"
  TRINO_CATALOG: "mongodb"
  TRINO_SCHEMA: "master"
  CHAT_SERVER: "https://bedrock-runtime.us-west-2.amazonaws.com"
  CHAT_SERVER_TYPE: "bedrock"
  CHAT_SERVER_MODEL: "qwen.qwen3-coder-480b-a35b-v1:0"
  NGINX_PORT: "5270"
 
---
 
apiVersion: v1
kind: ConfigMap
metadata:
  name: workbench2-nginx-config
  namespace: ecosystem
data:  
  supervisord.conf: |
    [supervisord]
    nodaemon=true
    logfile=/var/log/supervisor/supervisord.log
    pidfile=/tmp/supervisord.pid
    childlogdir=/var/log/supervisor
 
    [unix_http_server]
    file=/tmp/supervisor.sock
    chmod=0700
 
    [rpcinterface:supervisor]
    supervisor.rpcinterface_factory = supervisor.rpcinterface:make_main_rpcinterface
 
    [supervisorctl]
    serverurl=unix:///tmp/supervisor.sock
 
    # ==============================================================================
    # H2O Java Scorer
    # ==============================================================================
    [program:h2o-scorer]
    command=java -jar /app/h2o/h2o-scorer-1.0.0.jar --port %(ENV_H2O_SCORER_PORT)s --models %(ENV_H2O_MODELS_DIR)s
    directory=/app/h2o
    autostart=true
    autorestart=true
    startsecs=10
    stopwaitsecs=30
    stdout_logfile=/var/log/supervisor/h2o-scorer.log
    stderr_logfile=/var/log/supervisor/h2o-scorer-error.log
    stdout_logfile_maxbytes=50MB
    stderr_logfile_maxbytes=50MB
    environment=JAVA_OPTS="-Xmx2g -Xms512m"
    priority=100
 
    # ==============================================================================
    # Backend FastAPI
    # ==============================================================================
    [program:backend]
    command=python -m uvicorn app.main:app --host 0.0.0.0 --port %(ENV_BACKEND_PORT)s
    directory=/app/backend
    autostart=true
    autorestart=true
    startsecs=5
    stopwaitsecs=30
    stdout_logfile=/dev/stdout
    stdout_logfile_maxbytes=0
    stderr_logfile=/dev/stderr
    stderr_logfile_maxbytes=0
    environment=PYTHONPATH="/app/backend/src",PYTHONUNBUFFERED="1"
    priority=200
 
    # ==============================================================================
    # Nginx (Frontend)
    # ==============================================================================
    [program:nginx]
    command=/usr/sbin/nginx -g "daemon off;"
    autostart=true
    autorestart=true
    startsecs=5
    stopwaitsecs=10
    stdout_logfile=/var/log/supervisor/nginx.log
    stderr_logfile=/var/log/supervisor/nginx-error.log
    stdout_logfile_maxbytes=50MB
    stderr_logfile_maxbytes=50MB
    priority=300
 
    # ==============================================================================
    # Group all services
    # ==============================================================================
    [group:ecosystem]
    programs=h2o-scorer,backend,nginx
    priority=999
 
---
 
apiVersion: v1
kind: Secret
metadata:
  name: workbench2-secrets
  namespace: ecosystem
type: Opaque
stringData:
  CHAT_SERVER_KEY: "INSERTKEYHERE"
  VITE_GOOGLE_MAPS_API_KEY: "INSERTKEYHERE"
 
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: workbench2
  namespace: ecosystem
spec:
  replicas: 1
  selector:
    matchLabels:
      app: workbench2
  template:
    metadata:
      labels:
        app: workbench2
    spec:
      containers:
        - name: workbench2
          image: docker.io/ecosystemai/ecosystem-workbench2:amd64
          imagePullPolicy: Always
          ports:
            - name: http
              containerPort: 5270
          envFrom:
            - configMapRef:
                name: workbench2-config
            - secretRef:
                name: workbench2-secrets
          resources:
            requests:
              memory: 1Gi
            limits:
              memory: 2Gi
          volumeMounts:
            - name: nginx-config
              mountPath: "/etc/supervisor/conf.d/supervisord.conf"
              subPath: supervisord.conf
            - name: ecosystem-data
              subPath: data
              mountPath: /data
            - name: ecosystem-data
              subPath: workbench2-logs
              mountPath: /var/log/supervisor
          securityContext:
            runAsGroup: 0
      volumes:
        - name: nginx-config
          configMap:
            name: workbench2-nginx-config
        - name: ecosystem-data
          persistentVolumeClaim:
            claimName: ecosystem-data-pvc
✏️
Commands to Create Workbench2â–·
oc apply -f new-workbench.yaml
 
oc expose deployment workbench2 --port=5270
oc expose svc workbench2 --port=5270

Conclusion

That’s it! You have now configured your ecosystem.Ai instance on OpenShift. Superset and Airflow components can be added using helm charts if required.

Last updated on
KubernetesPost Install Setup Guide