OpenMetaData Deployment on Operate First#

The OpenMetaData (OM) deployment on Operate First leverages the Helm k8s deployment method. In order to stay consistent with our other services, the helm chart is converted to kustomize. This can be easily achieved by

helm repo add open-metadata

# Get dependencies
helm pull open-metadata/openmetadata-dependencies
tar -xvzf openmetadata-dependencies-${chart-version}>.tgz
helm template --name-template=openmetadata-dependencies .

# Get openmetadata
helm pull open-metadata/openmetadata
tar -xvzf openmetadata-${chart-version}.tgz
helm template --name-template=openmetadata .

Further additions/amendments are made to these renders. For the resulting manifests you can find them here.

Some examples of additions include:

  • Addition of OCP routes

  • Patches to deployments

  • Removal of any security context fields, or fields that attempt to change container uid

  • Affinity rules

Dependencies should be applied first, they consist of Elasticsearch, Mysql, and Airflow.

ReadWriteMany Access Mode#

The OpenMetaData Dependencies require ReadWriteMany Access Mode. These pods will attempt to bind the same pvcs for DAGs and Logs. Operate First Clusters do not have any storage providers that support this. However, we workaround this by ensuring all such pods are scheduled on the same node allowing them to treat a pvc with ReadWriteOnce accessmode as ReadWriteMany, see here for more info.


There are 2 access points that require authentication. They are as follows:

OpenMetaData UI#

This is configured using the Operate First OSC-Cl2 Dex Connector. Only the direct GitHub Authentication is supported, as of now. Authentication via Dex -> Openshift will not work (issue here). The OM documentation for this configuration can be found here. And the configuration is patched here.


Ingestion authentication is done through JWT tokens. Airflow is used to perform ingestion, and Airflow authenticates with OM via a JWT token that is generated via the OM UI.

First, for OM to be able to generate tokens, it needs JWT configuration:

This is documented here. You can find these fields in our configuration manifests here.

  rsapublicKeyFilePath: ${RSA_PUBLIC_KEY_FILE_PATH:-""}
  rsaprivateKeyFilePath: ${RSA_PRIVATE_KEY_FILE_PATH:-""}
  jwtissuer: ${JWT_ISSUER:-""}
  keyId: ${JWT_KEY_ID:-"Gb389a-9f76-gdjs-a92j-0242bk94356"}

Note the RSA_PUBLIC_KEY_FILE_PATH and RSA_PRIVATE_KEY_FILE_PATH, these are env variables made available in the deployment patch here. The values were generated using openssl:

openssl genrsa -out private_key.pem 2048
openssl pkcs8 -topk8 -inform PEM -outform DER -in private_key.pem -out private_key.der -nocrypt
openssl rsa -in private_key.pem -pubout -outform DER -out public_key.der

Then add these to a k8s secret for consumption by the OM pod. Once configured you should be able to generate a new token via OM ui.

The token is received from OM through the UI after logging in using the method as described above. Once done, the user needs to navigate to Settings > Bots > ingestion-bot and generate a new token from this ui. You will need to be using OM +0.11 to generate an infinite token, until then the longest expiry that can be selected is 7 days.

We set OM_AUTH_JWT_TOKEN in the OM pod here. We also set AIRFLOW_AUTH_PROVIDER as such:

  value: openmetadata
  value: <omitted>