Terraform Ops Manager

0 Disclaimer

These code samples are provided for educational and illustrative purposes only, to demonstrate the functionality of specific MongoDB features. They are not production-ready and may lack the security hardening, error handling, and testing required for a live environment.

You are responsible for testing, validating, and securing this code within your own environment before implementation. This material is provided "as is" without warranty or liability.’

1 What's Inside

The tool will create Ops Manager on AWS that has the following features:

• An Ops Manager instance
• A standalone AppDB
• A database for backup which has configurable number of members.
• Some EC2 instances that has automation running and pointing to "TestProject". You can use these instances to build your cluster for testing.
• One of the backup solutions configured:
  • S3
  • Oplog store + Blockstore (Both using the same DB)
  • Oplog store + FileSystem store

2 Structure of Repo

The main folders that you need to use are:

• 1-om: Terraform scripts that create:
  • Ops Manager application
  • AppDB
• 2-clusters: Terraform scripts that creates:
  • Test instances
  • Backup database (Oplog/Block store, S3 metadata store)
  • S3 buckets for backup (Oplog/Snapshot)

Due to the limitation of Terraform, I can't do everything in 1 phase. You need to finish 1-om then start 2-clusters.

3 Use the Tool

Be aware of the known issues. Some of them can fail the deployment.

3.1 Required Variables

I tried to make everything autowired so you only need to fill some necessary information. The other variables can let you customize the Ops Manager but they have defaults. All the important variables are described below. You can find in 1-om/variables.tf all the other variables and instructions.

aws_config

aws_config.region

Required
AWS region that you want to create the environment.

aws_config.vpc_id

Required
The VPC ID that you want to host the environmenbt. The VPC must exist and can assign public IP addess for the EC2 instances.

aws_config.sub_net

Required
The subnet where you want to create the EC2 instances. The subnet must exist.

aws_config.key_name

Required
The AWS key pair that will be assigned to all the EC2 instances. You can use the key to ssh the instances.

tags

tags.owner

Required
Your MongoDB email.

tags.expire-on

Optional
Use empty string auto fill a date that represents 3 days later. Otherwise please fill in the format of: yyyy-MM-dd.

tags.project-id

Optional
Defaults to internal. You have responsibility to fill the real PS project ID from Salesforce. The format is: PS-<project ID>.

backing_db_credentials

backing_db_credentials.name

Required
Name of user that will be created in all backing databases as root user. Including:

• AppDB
• Oplog/Block store
• S3 metadata store

backing_db_credentials.pwd

Required
Password for the user above.

first_user

The first Ops Manager user. This will be the Ops Manager admin.

first_user.email

Required
Email of the user.

first_user.pwd

Required
Password of the user.

first_user.fistName

Required
User's first name.

first_user.pwd

Required
User's last name.

3.2 Deploy Ops Manager

AWS Credentials

• Open AWS from company portal.
• Click "Access keys"
• Copy credential from "Option 1: Set AWS environment variables"
• Paste into bash

Execute Script

Create Ops Manager:

git clone https://github.com/zhangyaoxing/terraform-ops-manager
cd 1-om
terraform init
terraform apply

Create the rest:

cd 2-clusters
terraform init
terraform apply

Destroy the environment:

cd 2-clusters
terraform destroy
cd ../1-om
terraform destroy

4 Known Issues

4.1 Cloudflare

Cloudflare WARP is known to prevent us from sending big post requests. Thus the automation API requests in 2-clusters will fail.
Turn off Cloudflare WARP or use DoH mode will fix the problem.

If you have finished 1-om with WARP on, you need to go to Admin->Global Access List to whitelist your public IP so the script can work properly. Note the new whitelist takes 5 min to take effect.

4.2 Ops Manager API

There's no Terraform provider for Ops Manager. So I have to create some Python scripts to call the API, and use the null_resource to call the Python scripts. In this case, the resources (Orgs, Projects, Clusters) in Ops Manager will not be properly destroyed. This is usually harmless because they will be destroyed when you destroy Ops Manager.

4.3 Standalone AppDB

Although you can customize how many instances of backup DB, test DB you need, the AppDB currenly only support a single instance standalone. This should be enough for most use cases.

4.4 Certificates

Features that needs certificates are not supported at the moment. Including:

• HTTPS
• Queryable backup

4.5 Email

SMTP is not configured at the moment. All features that rely on the email will fail.