ElastiCache clusters should not use the default subnet group

How to fix it

• Replaces the resource: A cluster's subnet group cannot be changed, so create the group, back up the cluster, restore into a new cache in the group, repoint clients, then delete the old cluster.
• Can break clients: The replacement has a new endpoint for every client to repoint to, and a node-based Memcached cluster has no backups, so its replacement starts empty.

In the console

ElastiCache → Subnet groups → Create subnet group, then the Subnet group setting on the replacement cache

Name

Any name other than default

VPC ID

The VPC the replacement cache will run in

Selected subnets

Only the private subnets the cache needs

Subnet group

The new group, set when restoring or creating the replacement cache

With Terraform

resource "aws_elasticache_cluster" "this" {
  cluster_id        = var.cluster_id
  engine            = var.engine
  subnet_group_name = aws_elasticache_subnet_group.custom.name
}