OceanProtocol Provider on Kubernetes

Deploy OceanProtocol Provider on Kubernetes, for example on minikube.

First of all, to enable nginx ingress, run:

minikube addons enable ingress

To run locally, configure in /etc/hosts the following entry pointing to the IP of the minikube cluster, which can be obtained by running minikube ip:

192.168.64.4    ipfs.local provider.local

Then, follow the steps below.

Define dataspace namespace

kubectl create namespace dataspace

Deploy IPFS to persist Compute-to-Data outputs and logs

kubectl apply -f ipfs/ipfs-data-pvc.yaml 
kubectl apply -f ipfs/ipfs-deployment.yaml
kubectl apply -f ipfs/ipfs-service.yaml
kubectl apply -f ipfs/ipfs-ingress.yaml

Replace ipfs.local in ipfs-ingress.yaml with the hostname you are deploying the service to. Also replace the storage class in ipfs-data-pvc.yaml with the one you are using in your cluster.

Deploy Postgres to manage computation jobs requests

kubectl apply -f postgres/postgres-configmap.yaml
kubectl apply -f postgres/postgres-data-pvc.yaml 
kubectl apply -f postgres/postgres-deployment.yaml
kubectl apply -f postgres/postgres-service.yaml

Replace the storage class in postgres-data-pvc.yaml with the one you are using in your cluster.

Deploy the OceanProtocol Provider

kubectl apply -f provider/provider-deployment.yaml
kubectl apply -f provider/provider-service.yaml
kubectl apply -f provider/provider-ingress.yaml

Replace provider.local in provider-ingress.yaml with the hostname you are deploying the service to.

Moreover, don't forget to replace the PROVIDER_PRIVATE_KEY and UNIVERSAL_PRIVATE_KEY ENV vars in provider-deployment.yaml with private keys only known to you.

Deploy the OceanProtocol Operator Engine

kubectl apply -f operator-engine/operator-engine-service-account.yaml
kubectl apply -f operator-engine/operator-engine-role.yaml
kubectl apply -f operator-engine/operator-engine-role-binding.yaml
kubectl apply -f operator-engine/operator-engine-deployment.yaml

Update IPFS_OUTPUT_PREFIX and IPFS_LOG_PREFIX ENV vars in operator-engine-deployment.yaml with the hostname you deployed the IPFS service to. And use a STORAGE_CLASS that is available in your cluster and, preferable, that is automatically released and deleted after its use.

Moreover, don't forget to replace the OPERATOR_PRIVATE_KEY ENV var in operator-engine-deployment.yaml with a private key only known to you.

Deploy the OceanProtocol Operator API

kubectl apply -f operator-api/operator-api-deployment.yaml
kubectl apply -f operator-api/operator-api-service.yaml

Update the ALLOWED_PROVIDERS and OPERATOR_ADDRESS ENV vars in operator-api-deployment.yaml with the public keys of the Provider and Operator Engine you previously deployed.

Initialize Postgres

Finally, initialize the Postgres database through the Operator API from the Provider pod. Note that the Operator API is not visible externally. To do so, first open a shell in the Provider pod:

kubectl exec $(kubectl get pods -n dataspace -l=app=provider -o name) -n dataspace -it -- bash

From the shell in the Provider pod, first, install the curl command:

apt-get update && apt-get install curl -y

Then, call pgsqlinit in the Operator API using the Admin header to authenticate with one of the ALLOWED_ADMINS defined in the corresponding Operator API ENV var in operator-api-deployment.yaml):

curl -X POST "http://operator-api.dataspace:8050/api/v1/operator/pgsqlinit" -H  "accept: application/json" -H "Admin: postgresadmin"

Test Provider Availability

To check that the Provider is available, visit the Provider URL in your browser, for instance: http://provider.local

You should see the Provider API documentation:

{
  "chainIds": [
    100
  ],
  "providerAddresses": {
    "100": "0x7E5F4552091A69125d5DfCb7b8C2659029395Bdf"
  },
  "serviceEndpoints": {
    "computeDelete": [
      "DELETE",
      "/api/services/compute"
    ],
    "computeEnvironments": [
      "GET",
      "/api/services/computeEnvironments"
    ],
    "computeResult": [
      "GET",
      "/api/services/computeResult"
    ],
    "computeStart": [
      "POST",
      "/api/services/compute"
    ],
    "computeStatus": [
      "GET",
      "/api/services/compute"
    ],
    "computeStop": [
      "PUT",
      "/api/services/compute"
    ],
    "create_auth_token": [
      "GET",
      "/api/services/createAuthToken"
    ],
    "decrypt": [
      "POST",
      "/api/services/decrypt"
    ],
    "delete_auth_token": [
      "DELETE",
      "/api/services/deleteAuthToken"
    ],
    "download": [
      "GET",
      "/api/services/download"
    ],
    "encrypt": [
      "POST",
      "/api/services/encrypt"
    ],
    "fileinfo": [
      "POST",
      "/api/services/fileinfo"
    ],
    "initialize": [
      "GET",
      "/api/services/initialize"
    ],
    "initializeCompute": [
      "POST",
      "/api/services/initializeCompute"
    ],
    "nonce": [
      "GET",
      "/api/services/nonce"
    ],
    "validateContainer": [
      "POST",
      "/api/services/validateContainer"
    ]
  },
  "software": "Provider",
  "version": "2.1.6"
}

Which can be used to list the available compute environments: http://provider.local/api/services/computeEnvironments

Providing something like:

{
  "100": [
    {
      "consumerAddress": "0x7E5F4552091A69125d5DfCb7b8C2659029395Bdf",
      "cpuNumber": "2",
      "cpuType": "",
      "currentJobs": 0,
      "desc": "",
      "diskGB": "1",
      "feeToken": "0x0995527d3473b3a98c471f1ed8787acd77fbf009",
      "gpuNumber": "0",
      "gpuType": "",
      "id": "dataspace",
      "lastSeen": "1701725217.16205",
      "maxJobDuration": "900",
      "maxJobs": "1",
      "priceMin": "0.01",
      "ramGB": "4",
      "storageExpiry": 0
    }
  ]
}

rogargon / oceanprotocol-provider-kubectl