HA is achieved with deployments in two aws regions, namely:
- us-east-2 (primary)
- us-west-1 (DR)
In addition, the VPC has IPs in multiple availability zones, for the VMs and for the database.
| Asset | Purpose | Size | Qty | DR |
|---|---|---|---|---|
| Asset name | Brief description | AWS size eg. t3.micro (if applicable) | Number of nodes/replicas | Identify if this asset is deployed to DR, replicated, created in multiple locations or just stored elsewhere |
| -------------- | --------------------------------------------------- | --------------------------------------- | -------------------------- | -------------------------------------------------------------------------------------------------------------- |
| VM | web app with API | t3.micro | 3 instances (EC2) | Deployed to 2 regions with 2 availability zones each (access via load balancer) |
| Load Balancer | ALB balances the load between web app VMs | 2 | Application load balancer in each region | |
| EKS | Kubernetes cluster with monitoring stack | 2 nodes | 2 | Deployed to 2 regions with 2 availability zones each |
| RDS cluster | Database with 2 instance nodes for each cluster | db.t2.small | 2 (primary and secondary) | Replicated from zone1 to zone2 |
| EC2 keys | Access EC2 instances | 2 (one for each region) | ||
| -------------- | --------------------------------------------------- | --------------------------------------- | -------------------------- | -------------------------------------------------------------------------------------------------------------- |
More detailed descriptions of each asset identified above.
Steps we would perform to setup the infrastructure in the other region. Its not super detailed, but high-level should suffice.
Ensure the infrastructure is set up and working in the DR site.
Ensure both sites are configured the same
We won't actually perform these steps, but describe what we would do to "fail-over" our application and database cluster to the other region. We cover all the pieces that were setup and how we use those in the other region
- Create a cloud load balancer and point DNS to the load balancer.
- This way you can have multiple instances behind 1 IP in a region. During a failover scenario, you would fail over the single DNS entry at your DNS provider to point to the DR site.
- This is much more intelligent than pointing to a single instance of a web server.
- Have a replicated database and perform a failover on the database.
- While a backup is good and necessary, it is time-consuming to restore from backup. In this DR step, you would have already configured replication and would perform the database failover.
- Ideally, your application would be using a generic CNAME DNS record and would just connect to the DR instance of the database.
- Point your DNS to your secondary region
- This can be done with a name provider like Amazon route 53
- Failover your database replication instances to another region
- Manually force the secondary region to become primary at the database level, or
- Automatically failover the database by health checks