Deploy HStream on AWS with Terraform

This document describes how to deploy HStream on AWS with Terraform.

You can download related resources from hstreamdb/terraform:

git clone git@github.com:hstreamdb/terraform.git

Pre-Require

We assume that you have a basic understanding of terraform. Here are some resources may help:

• Terraform Language Documentation (opens new window)
• Terraform tutorials (opens new window)

Also, you need to have an AWS account and create your own key-pair so that you can use it to access AWS EC2 instance later.

Install Terraform

Refer to terraform install doc (opens new window)

Configuration

Let's take a look at the structure of the project first.

| - file
     | - clusterCfg.json.tftpl // cluster configuration file
       - logdevice.conf        // hstore configuration file
       - prometheus-cfg        // prometheus related config files
       ...
  - main.tf                    // terraform main program
  - start.py                   // a script use to start HStream cluster
  - terraform.tfvars           // external variables used by terraform
  ...

Listed above are the key components and associated configuration files to start the cluster, which you need to configure very carefully.

Configure cluster-related information

You should modify file/clusterCfg.json.tftpl to configure your own HStream cluster. You can refer to Create a Configuration File (opens new window) to complete the configuration file. Notice: There are two places that differ from the part described in the link：

• You don't need to fill in the hosts filed, this part will be done by terraform.

• Refer to the corresponding server nodes according to the following rules:

  • Nodes can be divided into two categories：
    • hs-s*：This class of nodes uses AWS storage-optimized instances configured with NVMe SSDs and is mainly used to boot store related services.
    • hs-c*: This class of nodes are typically used to boot compute related services.
  • The * can only be a number

• As an example, in our default configuration, we start 3 hs-s* nodes and 1 hs-c* node:

  nodes	Amazon EC2 Instance Types	instance
  hs-s1	i3en.2xlarge	hstore、hserver、zookeeper、node-exporter、cadvisor
  hs-s2	i3en.2xlarge	hstore、hserver、zookeeper、node-exporter、cadvisor
  hs-s3	i3en.2xlarge	hstore、hserver、zookeeper、node-exporter、cadvisor
  hs-c1	c5.4xlarge	prometheus、hstore-admin、node-exporter、cadvisor

Configure hstore

Refer to Create a configuration file (opens new window) to modifyfile/logdevice.conf. Also, no need to modify the zookeeper_uri field, the program will modify it automatically.

Configure terraform

Modify terraform.tfvars to config terraform

• region：The instance's region
• image_id：The system image you want to use, we choose ubuntu 20.04 here
• delete_block_on_termination：Whether to delete block devices at the same time when deleting instances
• cidr_block: Specific VPC
• private_key_path：The private key's path to access AWS instance
• key_pair_name：The key-pair's name you created before
• store_config and cal_config: We divide all AWS instances into two categories: one that performs primarily storage tasks, which we call "storage nodes", and another that performs primarily compute tasks, which we call "compute nodes". Each node has the following attributes：
  • node_count：The count of node instances
  • instance_type: Amazon EC2 Instance Types
  • volume_type: The type of storage volume
    • volume_size、iops、throughput are attributes of storage volume

Set Up the Cluster

Now, we can set up our cluster.

First, you should export AWS_SECRET_ACCESS_KEY

export AWS_ACCESS_KEY_ID="<Your access key id>"
export AWS_SECRET_ACCESS_KEY="<Your secret access key>"

Then you can use ./start.py -k <your private key's path> to set up the cluster.