Introduction

Let me start saying that terraform is an amazing tool for infrastructure automation and gives you the ability to be Truly Cloud Agnostic and manage a state for your infrastructure I have been working with terraform for quite sometime and have used pretty most of the plugins and providers.

The Problem: Restoring ElastiCache from Snapshots

Probably a year back I was working on module for Amazon ElastiCache and stumbled upon a issue: If you want to create a elasticache cluster from backups to the mentioned slots using terraform then terraform gives up!!

As the plugin doesn’t support that.

Limitations of the Terraform AWS Provider

Though it is pretty much possible from Amazon UI, means APIs are there however the terraform resource elasticache_replication_group doesn’t support that.

The variable snapshot_arns can be used to provide the location of the backups, which is usually S3, however there is not way to assign Slots for the them I decided to check the elasticache_replication_group code and see if it is possible to introduce the same.

I looked at the code As you can see that snapshot_arns are there but no way to provide Slots information. There is Github Issue for this as well

The Solution: Custom Fork and Enhancement

As a result I forked the repo as tfproviders/aws-terraform-provider and introduced following changes: Add node_group_configuration {} block node_group_configuration takes slots as list(string) So now my terraform code for aws_elasticache_replication_group looks like this

resource "aws_elasticache_replication_group" "restore" {
  replication_group_id = "tf-redis-cluster"
  replication_group_description = "test description"
  node_type = "cache.t2.small"
  port = 6379
  parameter_group_name = "default.redis3.2.cluster.on"
  automatic_failover_enabled = true
  snapshot_arns = [ "arn:aws:s3:::s3-bucket/snapshot-0001.rdb", "arn:aws:s3:::s3-bucket/snapshot-0002.rdb"]
  node_group_configuration {
    slots = [ "0-5461", "5462-10922" ]
  }
  cluster_mode {
    replicas_per_node_group = 1
    num_node_groups = 2
  }
}

Result

Each element of Slots maps to snaphot_arns and Voila!!

I was able to create Amazon ElastiCache using terraform resource elasticache_replication_group from backups with slots information.

Usage

To use my forked version of Terraform AWS Provider add following to your terraform code and here is the published provider tfproviders/aws.

terraform {
  required_providers {
    aws = {
      source = "tfproviders/aws"
      version = "4.66.0"
    }
  }
}
provider "aws" {
  # Configuration options
}

Conclusion

Happy Terraforming!!