This is a quickstart guide for deploying TigerGraph in Kubernetes on AWS’s Elastic Kubernetes Service (EKS).

 Known issue: The GUI service, which is the service for GraphStudio and AdminPortal UI, does not currently support high availability in an EKS cluster.

## 1. Before you begin

• Provision Kubernetes cluster on EKS with nodes that meet the hardware and software requirements to run TigerGraph.

• Install `kubectl` on your machine, and make sure your local `kubectl` version is within one minor version’s difference from the `kubectl` version on your cluster.

• Configure `kubectl` for EKS cluster access.

• Create and delete Pods, Services, StatefulSets, and ConfigMaps

• Create and delete Jobs, CronJobs

• Create and delete Service Accounts, roles and role bindings

 Each of the commands below uses `kubectl` with the default namespace `default`. If you’ve deployed your cluster using a different namespace, you need to explicitly provide the namespace where your cluster is deployed.

## 2. Single-server deployment

This section describes the steps to deploy, verify, and remove a single-server deployment of TigerGraph on EKS.

### 2.1. Deploy single server

#### 2.1.1. Generate deployment manifest

Clone the TigerGraph ecosystems repository and change into the `k8s` directory. You can edit the `kustimization.yaml` file in the `eks` folder to change the namespace and image name for your deployment. The default namespace is `default`. No need to edit the files if no changes are needed.

Next, run the `./tg` script in the `k8s` directory to generate the deployment manifest for a single-server deployment. You can use the `--prefix` option to specify a prefix for your pods. The default prefix is `tigergraph`. A `deploy` directory will be created automatically, and you should find the manifest named `tigergraph-eks-default.yaml` in the directory.

### 2.2. Verify single server

Run `kubectl get pods` to confirm that the pods were created successfully:

``````$kubectl get pods NAME READY STATUS RESTARTS AGE installer-zsnb4 1/1 Running 0 4m11s tigergraph-0 1/1 Running 0 4m10s`````` Run `kubectl get services` to find the IP addresses of the RESTPP service as well as the GUI service. You can then make curl calls to the IP address of `tg-rest-service` at port 9000 to make sure that RESTPP is running: ``````$ curl <restpp_ip>:9000/echo | jq .
% Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
100    39  100    39    0     0    120      0 --:--:-- --:--:-- --:--:--   120
{
"error": false,
"message": "Hello GSQL"
}``````

You can also copy the IP address of the GUI service into your browser and visit port 14240 to make sure that GraphStudio is working.

### 2.3. Connect to single server

You can use `kubectl` to get a shell to the container or log in via `ssh`

``````# Via kubectl
kubectl exec -it tigergraph-0 -- /bin/bash

# Via ssh
ip_m1=$(kubectl get pod -o wide |grep tigergraph-0| awk '{print$6}')
ssh tigergraph@ip_m1``````

### 2.4. Remove single server resources

Use the `tg` script in the `k8s` directory of the repo to delete all cluster resources. Replace `<namespace_name>` with the name of the namespace within which you want to delete the resources. If you don’t specify a namespace, the command will delete the resources in the namespace `default`:

``$./tg eks delete -n <namespace_name>`` ## 3. Cluster deployment Once your EKS cluster is ready, you can start following the below steps to deploy a TigerGraph cluster on Kubernetes. ### 3.1. Deploy TigerGraph cluster #### 3.1.1. Generate Kubernetes manifest Clone the TigerGraph ecosystem repository and change into the `k8s` directory: ``````$ git clone https://github.com/tigergraph/ecosys.git
$cd ecosys/k8s`````` You can customize your deployment by editing the `kustomize.yaml` file in the `eks` directory. The `tg` script in the `k8s` folder offers a convenient way to make common customizations such as namespace, TigerGraph version, as well as cluster size. Run `./tg -h` to view the help text on how to use the script. Use the `tg` script in the `k8s` directory of the repo to create a Kubernetes manifest. Use `-s` or `--size` to indicate the number of nodes in the cluster. Use the `--ha` option to indicate the replication factor of the cluster, and the partitioning factor will be the number of nodes divided by the replication factor. For example, the following command will create a manifest that will deploy a 3*2 cluster with a replication factor of 2 and a partitioning factor of 3. ``$ ./tg eks kustomize -s 6 --ha 2``

The command will create a directory named `deployment` with the manifest inside.

#### 3.1.2. Deploy the cluster

Run `kubectl apply` to create the deployment

``$kubectl apply -f ./deploy/<namespace>-eks/tigergraph-eks-default.yaml`` ### 3.2. Verify cluster Run `kubectl get pods` to verify the pods were created successfully: ``````$ kubectl get pods
Run `kubectl get services` to find the IP addresses of the RESTPP service as well as the GUI service. You can then make curl calls to the IP address of `tg-rest-service` at port 9000 to make sure that RESTPP is running: