CloudFormation Template Generator

This page shows how to create CloudFormation templates using an API call. The same API calls are used by the Self-Service Portal to generate the CloudFormation template before redirecting to AWS.

Before Starting

The APIs described here require an API token which can be obtained at Obtaining this token requires registration if you do not have an account.

API Overview

To generate a CloudFormation template, it is first necessary to decide which Weka system version is to be installed. This is performed using the https://<token> API which provides a list of all available versions:

$ curl https://<token>
"num_results" : 8,
"page" : 1,
"page_size" : 50,
"num_pages" : 1,
"objects" : [
"id" : "3.6.1",
"public" : true,
"final" : true,
"trunk_id" : "",
"s3_path" : "releases/3.6.1"

This list of releases available for installation is sorted backward from the most recent release. By default, 50 results are provided per page. To receive more results, use the page=N query parameter to receive the Nth page.

Note: Usually, a request from more results is not necessary, since the first page contains the most recent releases.

Each release contains an ID field that identifies the release. In the examples below, version 3.6.1 has been used.

To generate a CloudFormation template, make a POST request to the https://<token><version>API:

$ spec='
"cluster": [
"role": "backend",
"instance_type": "i3en.2xlarge",
"count": 10
"role": "client",
"instance_type": "r3.xlarge",
"count": 2
$ curl -X POST -H 'Content-Type: application/json' -d "$spec" https://<token>
"url" : "",
"quick_create_stack" : {
"ap-southeast-2" : "...",

In the example above, a template was generated for a cluster with 10 i3en.2xlarge backend instances and 2 r3.xlarge client instances. Refer to the Deployment Types page to learn more, and see all supported instance types in Supported EC2 Instance Types.

Request Body

The https://<token><version> API provides a JSON object with a cluster property. cluster is a list of instance types, roles, and counts:




Either backend or client;see Deployment Types for more information.


One of the supported instance types according to the role and supported instances in Supported EC2 Instance Types.


The number of instances of this type to be included in the template.


When role is client, it is possible to specify a custom AMI-ID. See Custom Client AMI below to learn more.


Either dedicated or shared, in client role only. See Dedicated vs. Shared Client Networking below.

It is possible to specify multiple groups of instances by adding more role/instance_type/count objects to the clusterarray, as long as there are at least 6 backend instances (the minimum number of backend instances required to deploy a cluster).

Custom Client AMI

When specifying an ami_id in client groups, the specified AMI will be used when launching the client instances. The Weka system will be installed on top of this AMI in a separate EBS volume.

When ami_id is not specified, the client instances are launched with the latest Amazon Linux supported by the Weka system version selected to be installed.

Note the following when using a custom AMI-ID:

  • AMIs are stored per region. Make sure to specify an AMI-ID that matches the region in which the CloudFormation template is being deployed.

  • The AMI operating system must be one of the supported operating systems listed in the prerequisites page of the version being installed. If the AMI defined is not supported or has an unsupported operating system, the installation may fail and the CloudFormation stack will not be created successfully.

Dedicated vs. Shared Client Networking

By default, both client and backend instances are launched in the dedicated networking mode. Although this cannot be changed for backends, it can be controlled for client instances.

Dedicated networking means that an ENI is created for internal cluster traffic in the client instances. This allows the Weka system to bypass the kernel and provide throughput that is only limited by the instance network.

In shared networking, the client shares the instance’s network interface with all traffic passing through the kernel. Although slower, this mode is sometimes desirable when an ENI cannot be allocated or if the operating system does not allow more than one NIC.

Returned Result

The returned result is a JSON object with two properties: url and quick_create_stack.

The url property is a URL to an S3 object containing the generated template.

To deploy the CloudFormation template through the AWS console, a quick_create_stack property contains links to the console for each public AWS region. These links are pre-filled with your API token as a parameter to the template.

Note: CloudFormation template URLs are valid for up to 1 week.

It is also possible to receive the template directly from the API call, without saving it in a bucket. To do this, use a ?type=templatequery parameter:

$ spec='...' # same as above
$ curl -X POST -H 'Content-Type: application/json' -d "$spec" https://<token>
{"AWSTemplateFormatVersion": "2010-09-09", ...

CloudFormation Template Parameters

The CloudFormation template has the following parameters:




SSH key for the ec2-user that will be used to connect to the instances.


VPC in which the Weka cluster will be deployed.


The subnet in which the Weka cluster will be deployed.


Load balancer type for serving the cluster UI:

  • Internet Facing sets up the load balancer with a public IP.

  • Internal sets up the load balancer with a private IP address in the selected subnet.

  • No Load Balancer skips load balancer creation, in which case the UI can be accessed through port 14000 of any of the backend instances.

Weka uses a Classic Load Balancer (CLB), which is not supported in all zones/regions (e.g., not available in local zones). If CLB is not available in your zone/region, select "No Load Balancer".


Volume type for the Weka partition. GP3 is not yet available in all zones/regions (e.g., not available in local zones). In such a case, you must select the GP2 volume type. When available, using GP3 is preferred.


API token for Weka's distribution site. This can be obtained at


Sets the admin password after the cluster has been created. If no value is provided, the password is set to admin.


New S3 bucket name to create and attach to the filesystem created by the template. The bucket will not be deleted when the stack is destroyed.


Existing S3 bucket name to attach to the filesystem created by the template. The bucket has to be in the same region where the cluster is deployed. If this parameter is provided, the New S3 Bucket parameter is ignored.


Sets how much of the filesystem capacity (in percent) should reside on SSD. This parameter is applicable only if New S3 Bucket or Existing S3 Bucket parameters have been defined.

Note: It is possible using a NAT (with LoadBalancerType parameter other than Internet Facing). Otherwise,SubnetId must be a public subnet or a subnet that can be routed to the Internet, since deployment is based on downloading the Weka system and various packages from the Internet.

IAM Role Created in the Template

The CloudFormation template contains an instance role that allows the Weka system instances to call the following AWS APIs:

  • ec2:DescribeInstances

  • ec2:DescribeNetworkInterfaces

  • ec2:AttachNetworkInterface

  • ec2:CreateNetworkInterface

  • ec2:ModifyNetworkInterfaceAttribute

  • ec2:DeleteNetworkInterface

In case tiering is configured, additional AWS APIs permissions are given:

  • s3:DeleteObject

  • s3:GetObject

  • s3:PutObject

  • s3:HeadBucket

  • s3:ListBucket

Additional Operations

Once a CloudFormation template has been generated, it is possible to create a stack from it using the AWS console or the AWS CLI.

When the deployment is complete, the stack status will update to CREATE_COMPLETE and it is possible to access the Weka system cluster GUI by going to the Outputs tab of the CloudFormation stack and clicking the GUI link.


As part of the deployment, a filesystem is created and mounted on all instances. This shared filesystem is mounted on /mnt/weka in each of the cluster instances.

If the deployment was unsuccessful, see Troubleshooting for the resolution of common deployment issues.