Deploying a Pulsar cluster on AWS using Terraform and Ansible

For instructions on deploying a single Pulsar cluster manually rather than using Terraform and Ansible, see Deploying a Pulsar cluster on bare metal. For instructions on manually deploying a multi-cluster Pulsar instance, see Deploying a Pulsar instance on bare metal.

One of the easiest ways to get a Pulsar cluster running on Amazon Web Services (AWS) is to use the the Terraform infrastructure provisioning tool and the Ansible server automation tool. Terraform can create the resources necessary to run the Pulsar cluster—-EC2 instances, networking and security infrastructure, etc.—-while Ansible can install and run Pulsar on the provisioned resources.

Requirements and setup

In order install a Pulsar cluster on AWS using Terraform and Ansible, you'll need:

• An AWS account and the aws command-line tool
• Python and pip
• The terraform-inventory tool, which enables Ansible to use Terraform artifactsYou'll also need to make sure that you're currently logged into your AWS account via the aws tool:

$ aws configure

Installation

You can install Ansible on Linux or macOS using pip.

$ pip install ansible

You can install Terraform using the instructions here.

You'll also need to have the Terraform and Ansible configurations for Pulsar locally on your machine. They're contained in Pulsar's GitHub repository, which you can fetch using Git:

$ git clone https://github.com/apache/pulsar
$ cd pulsar/deployment/terraform-ansible/aws

SSH setup

If you already have an SSH key and would like to use it, you skip generating the SSH keys and update private_key_file settingin ansible.cfg file and public_key_path setting in terraform.tfvars file.

For example, if you already had a private SSH key in ~/.ssh/pulsar_aws and a public key in ~/.ssh/pulsar_aws.pub,you can do followings:

1. update ansible.cfg with following values:

private_key_file=~/.ssh/pulsar_aws

2. update terraform.tfvars with following values:

public_key_path=~/.ssh/pulsar_aws.pub

In order to create the necessary AWS resources using Terraform, you'll need to create an SSH key. To create a private SSH key in ~/.ssh/id_rsa and a public key in ~/.ssh/id_rsa.pub:

$ ssh-keygen -t rsa

Do not enter a passphrase (hit Enter when prompted instead). To verify that a key has been created:

$ ls ~/.ssh
id_rsa               id_rsa.pub

Creating AWS resources using Terraform

To get started building AWS resources with Terraform, you'll need to install all Terraform dependencies:

$ terraform init
# This will create a .terraform folder

Once you've done that, you can apply the default Terraform configuration:

$ terraform apply

You should then see this prompt:

Do you want to perform these actions?
  Terraform will perform the actions described above.
  Only 'yes' will be accepted to approve.
  Enter a value:

Type yes and hit Enter. Applying the configuration could take several minutes. When it's finished, you should see Apply complete! along with some other information, including the number of resources created.

Applying a non-default configuration

You can apply a non-default Terraform configuration by changing the values in the terraform.tfvars file. The following variables are available:

What is installed

When you run the Ansible playbook, the following AWS resources will be used:

• 9 total Elastic Compute Cloud (EC2) instances running the ami-9fa343e7 Amazon Machine Image (AMI), which runs Red Hat Enterprise Linux (RHEL) 7.4. By default, that includes:
  • 3 small VMs for ZooKeeper (t2.small instances)
  • 3 larger VMs for BookKeeper bookies (i3.xlarge instances)
  • 2 larger VMs for Pulsar brokers (c5.2xlarge instances)
  • 1 larger VMs for Pulsar proxy (c5.2xlarge instances)
• An EC2 security group
• A virtual private cloud (VPC) for security
• An API Gateway for connections from the outside world
• A route table for the Pulsar cluster's VPC
• A subnet for the VPCAll EC2 instances for the cluster will run in the us-west-2 region.

Fetching your Pulsar connection URL

When you apply the Terraform configuration by running terraform apply, Terraform will output a value for the pulsar_service_url. It should look something like this:

pulsar://pulsar-elb-1800761694.us-west-2.elb.amazonaws.com:6650

You can fetch that value at any time by running terraform output pulsar_service_url or parsing the terraform.tstate file (which is JSON, even though the filename doesn't reflect that):

$ cat terraform.tfstate | jq .modules[0].outputs.pulsar_service_url.value

Destroying your cluster

At any point, you can destroy all AWS resources associated with your cluster using Terraform's destroy command:

$ terraform destroy

Setup Disks

Before you run the Pulsar playbook, you want to mount the disks to the correct directories on those bookie nodes.Since different type of machines would have different disk layout, if you change the instance_types in your terraformconfig, you need to update the task defined in setup-disk.yaml file.

To setup disks on bookie nodes, use this command:

$ ansible-playbook \
  --user='ec2-user' \
  --inventory=`which terraform-inventory` \
  setup-disk.yaml

After running this command, the disks will be mounted under /mnt/journal as journal disk, and /mnt/storage as ledger disk.It is important to run this command only once! If you attempt to run this command again after you have run Pulsar playbook,it might be potentially erase your disks again and cause the bookies to fail to start up.

Running the Pulsar playbook

Once you've created the necessary AWS resources using Terraform, you can install and run Pulsar on the Terraform-created EC2 instances using Ansible. To do so, use this command:

$ ansible-playbook \
  --user='ec2-user' \
  --inventory=`which terraform-inventory` \
  ../deploy-pulsar.yaml

If you've created a private SSH key at a location different from ~/.ssh/id_rsa, you can specify the different location using the —private-key flag:

$ ansible-playbook \
  --user='ec2-user' \
  --inventory=`which terraform-inventory` \
  --private-key="~/.ssh/some-non-default-key" \
  ../deploy-pulsar.yaml

Accessing the cluster

You can now access your running Pulsar using the unique Pulsar connection URL for your cluster, which you can obtain using the instructions above.

For a quick demonstration of accessing the cluster, we can use the Python client for Pulsar and the Python shell. First, install the Pulsar Python module using pip:

$ pip install pulsar-client

Now, open up the Python shell using the python command:

$ python

Once in the shell, run the following:

>>> import pulsar
>>> client = pulsar.Client('pulsar://pulsar-elb-1800761694.us-west-2.elb.amazonaws.com:6650')
# Make sure to use your connection URL
>>> producer = client.create_producer('persistent://public/default/test-topic')
>>> producer.send('Hello world')
>>> client.close()

If all of these commands are successful, your cluster can now be used by Pulsar clients!