Running Marquez on AWS

This guide helps you deploy and manage Marquez on AWS EKS.

PREREQUISITES

AWS EKS Cluster

To create an AWS EKS cluster, please follow the steps outlined in the AWS EKS documentation.

CONNECT TO AWS EKS CLUSTER

Make sure you have configured your AWS CLI, then create or update the kubeconfig file for your cluster:
```
$ aws eks --region <AWS-REGION> update-kubeconfig --name <AWS-EKS-CLUSTER>
```

Verify that the context has been switched:

$ kubectl config current-context
arn:aws:eks:<AWS-REGION>:<AWS-ACCOUNT-ID>:cluster/<AWS-EKS-CLUSTER>

Using kubectl, verify that you can connect to your cluster:
```
$ kubectl get svc
NAME             TYPE        CLUSTER-IP   EXTERNAL-IP   PORT(S)   AGE
svc/kubernetes   ClusterIP   10.100.0.1   <none>        443/TCP   1m
```
Note: If you’re having issues connecting to your cluster, please see Why can’t I connect to my AWS EKS cluster?

AWS RDS

Next, create an AWS RDS instance as outlined in the AWS RDS documentation. This database will be used to store dataset, job, and run metadata collected as OpenLineage events via the Marquez HTTP API.

CREATE AWS RDS DATABASE

Navigate to the AWS RDS page and create a PostgreSQL database, leaving the database template as Production.
Use marquez as the database identifier and set the master username to marquez.
Choose a master password to use later in your Helm deployment (see password in values.yaml).
Leave public access to the database off.
Choose the same VPC where your AWS EKS cluster resides.
In a separate tab, navigate to the AWS EKS cluster page and make note of the security group attached to your cluster.
Navigate back to the AWS RDS page and, in the security group section, add the AWS EKS cluster’s security group from step 6.
Next, under the Additional Configuration tab, enter marquez as the initial database name.
Finally, select Create Database.

CONNECT TO AWS RDS DATABASE

Create a marquez namespace:
```
$ kubectl create namespace marquez
```

Next, run the following command with your AWS RDS host, username, and password:

kubectl run pgsql-postgresql-client --rm --tty -i --restart='Never' \
  --namespace marquez \
  --image docker.io/bitnami/postgresql:12-debian-10 \
  --env="PGPASSWORD=<AWS-RDS-PASSWORD>" \
  --command -- psql marquez --host <AWS-RDS-HOST> -U <AWS-RDS-USERNAME> -d marquez -p 5432

Deploy Marquez on AWS EKS

INSTALLING MARQUEZ

Get Marquez:

$ git clone git@github.com:MarquezProject/marquez.git && cd chart

Install Marquez:

helm upgrade --install marquez .
  --set marquez.db.host=<AWS-RDS-HOST>
  --set marquez.db.user=<AWS-RDS-USERNAME>
  --set marquez.db.password=<AWS-RDS-PASSWORD>
  --namespace marquez
  --atomic
  --wait

Note: To avoid overriding deployment settings via the command line, update the marquez.db section of the Marquez Helm chart’s values.yaml to include the AWS RDS host, username, and password in your deployment.

Verify all the pods have come up correctly:
```
$ kubectl get pods --namespace marquez
```

UNINSTALLING MARQUEZ

helm uninstall marquez --namespace marquez