This repo provides the Kubernetes Helm charts for deploying Datahub and it's dependencies (Elasticsearch, optionally Neo4j, MySQL, and Kafka) on a Kubernetes cluster.
- Set up a kubernetes cluster
- In a cloud platform of choice like Amazon EKS, Google Kubernetes Engine, and Azure Kubernetes Service OR
- In local environment using Minikube. Note, more than 7GB of RAM is required to run Datahub and it's dependencies
- Install the following tools:
Datahub consists of 4 main components: GMS, MAE Consumer (optional), MCE Consumer (optional), and Frontend. Kubernetes deployment for each of the components are defined as subcharts under the main Datahub helm chart.
The main components are powered by 4 external dependencies:
- Kafka
- Local DB (MySQL, Postgres, MariaDB)
- Search Index (Elasticsearch)
- Graph Index (Supports either Neo4j or Elasticsearch)
The dependencies must be deployed before deploying Datahub. We created a separate
chart
for deploying the dependencies with example configuration. They could also be deployed
separately on-prem or leveraged as managed services. To remove your dependency on Neo4j,
set enabled to false in the datahub-kubernetes/prerequisites/values.yaml
file.
Then, override the graph_service_impl
field in datahub-kubernetes/datahub/values.yaml
to
have the value elasticsearch
instead of neo4j
.
Assuming kubectl context points to the correct kubernetes cluster, first create kubernetes secrets that contain MySQL and Neo4j passwords.
kubectl create secret generic mysql-secrets --from-literal=mysql-root-password=datahub
kubectl create secret generic neo4j-secrets --from-literal=neo4j-password=datahub
The above commands sets the passwords to "datahub" as an example. Change to any password of choice.
Add datahub helm repo by running the following
helm repo add datahub https://helm.datahubproject.io/
Then, deploy the dependencies by running the following
helm install prerequisites datahub/datahub-prerequisites
Note, the above uses the default configuration defined here. You can change any of the configuration and deploy by running the following command.
helm install prerequisites datahub/datahub-prerequisites --values <<path-to-values-file>>
Run kubectl get pods
to check whether all the pods for the dependencies are running.
You should get a result similar to below.
NAME READY STATUS RESTARTS AGE
elasticsearch-master-0 1/1 Running 0 62m
elasticsearch-master-1 1/1 Running 0 62m
elasticsearch-master-2 1/1 Running 0 62m
prerequisites-cp-schema-registry-cf79bfccf-kvjtv 2/2 Running 1 63m
prerequisites-kafka-0 1/1 Running 2 62m
prerequisites-mysql-0 1/1 Running 1 62m
prerequisites-neo4j-community-0 1/1 Running 0 52m
prerequisites-zookeeper-0 1/1 Running 0 62m
deploy Datahub by running the following
helm install datahub datahub/datahub
Values in values.yaml have been preset to point to the dependencies deployed using the prerequisites chart with release name "prerequisites". If you deployed the helm chart using a different release name, update the quickstart-values.yaml file accordingly before installing.
Run kubectl get pods
to check whether all the datahub pods are running. You should get a result similar to below.
NAME READY STATUS RESTARTS AGE
datahub-acryl-datahub-actions-58b676f77c-c6pfx 1/1 Running 0 4m2s
datahub-datahub-frontend-84c58df9f7-5bgwx 1/1 Running 0 4m2s
datahub-datahub-gms-58b676f77c-c6pfx 1/1 Running 0 4m2s
datahub-datahub-mae-consumer-7b98bf65d-tjbwx 1/1 Running 0 4m3s
datahub-datahub-mce-consumer-8c57d8587-vjv9m 1/1 Running 0 4m2s
datahub-elasticsearch-setup-job-8dz6b 0/1 Completed 0 4m50s
datahub-kafka-setup-job-6blcj 0/1 Completed 0 4m40s
datahub-mysql-setup-job-b57kc 0/1 Completed 0 4m7s
elasticsearch-master-0 1/1 Running 0 97m
elasticsearch-master-1 1/1 Running 0 97m
elasticsearch-master-2 1/1 Running 0 97m
prerequisites-cp-schema-registry-cf79bfccf-kvjtv 2/2 Running 1 99m
prerequisites-kafka-0 1/1 Running 2 97m
prerequisites-mysql-0 1/1 Running 1 97m
prerequisites-neo4j-community-0 1/1 Running 0 88m
prerequisites-zookeeper-0 1/1 Running 0 97m
You can run the following to expose the frontend locally. Note, you can find the pod name using the command above.
In this case, the datahub-frontend pod name was datahub-datahub-frontend-84c58df9f7-5bgwx
.
kubectl port-forward <datahub-frontend pod name> 9002:9002
You should be able to access the frontend via http://localhost:9002.
Once you confirm that the pods are running well, you can set up ingress for datahub-frontend to expose the 9002 port to the public.
We welcome contributions from the community. Please refer to our Contributing Guidelines for more details.
Join our slack workspace for discussions and important announcements.