Enable cache to use cluster mode with node groups

elasticache-redis/replication-group/README.md

@@ -53,9 +53,11 @@ Provision a Redis cluster using AWS ElastiCache.
 | <a name="input_kms_key"></a> [kms\_key](#input\_kms\_key) | Custom KMS key to encrypt data at rest | `object({ arn = string })` | `null` | no |
 | <a name="input_name"></a> [name](#input\_name) | Name for this cluster | `string` | n/a | yes |
 | <a name="input_node_type"></a> [node\_type](#input\_node\_type) | Node type for the Elasticache instance | `string` | n/a | yes |
+| <a name="input_num_node_groups"></a> [num\_node\_groups](#input\_num\_node\_groups) | Number of node groups (shards) to create when sharding is enabled | `number` | `null` | no |
 | <a name="input_parameter_group_name"></a> [parameter\_group\_name](#input\_parameter\_group\_name) | Parameter group name for the Redis cluster | `string` | `null` | no |
 | <a name="input_port"></a> [port](#input\_port) | Port on which to listen | `number` | `6379` | no |
 | <a name="input_replica_count"></a> [replica\_count](#input\_replica\_count) | Number of read-only replicas to add to the cluster | `number` | `1` | no |
+| <a name="input_replicas_per_node_group"></a> [replicas\_per\_node\_group](#input\_replicas\_per\_node\_group) | Number of replicas to create in each node group when num\_node\_groups is set | `number` | `null` | no |
 | <a name="input_replication_group_id"></a> [replication\_group\_id](#input\_replication\_group\_id) | Override the ID for the replication group | `string` | `""` | no |
 | <a name="input_server_security_group_ids"></a> [server\_security\_group\_ids](#input\_server\_security\_group\_ids) | IDs of VPC security groups for this instance. One of vpc\_id or server\_security\_group\_ids is required | `list(string)` | `[]` | no |
 | <a name="input_server_security_group_name"></a> [server\_security\_group\_name](#input\_server\_security\_group\_name) | Override the name for the security group; defaults to identifer | `string` | `""` | no |

elasticache-redis/replication-group/main.tf

@@ -8,9 +8,11 @@ resource "aws_elasticache_replication_group" "this" {
   kms_key_id                 = var.kms_key == null ? null : var.kms_key.id
   multi_az_enabled           = local.replica_enabled
   node_type                  = var.node_type
-  num_cache_clusters         = local.instance_count
+  num_cache_clusters         = local.use_num_node_groups ? null : local.instance_count
+  num_node_groups            = local.num_node_groups
   parameter_group_name       = var.parameter_group_name
   port                       = var.port
+  replicas_per_node_group    = local.replicas_per_node_group
   description                = var.description
   security_group_ids         = local.server_security_group_ids
   snapshot_name              = var.snapshot_name
@@ -46,6 +48,16 @@ resource "aws_elasticache_replication_group" "this" {
       # Minor upgrades will cause noise in diffs
       engine_version
     ]
+
+    precondition {
+      condition     = !local.use_num_node_groups || var.replicas_per_node_group != null || ((local.instance_count % var.num_node_groups) == 0)
+      error_message = "When num_node_groups is set without replicas_per_node_group, replica_count + 1 must divide evenly across node groups."
+    }
+
+    precondition {
+      condition     = var.replicas_per_node_group == null || (local.instance_count == (var.num_node_groups * (var.replicas_per_node_group + 1)))
+      error_message = "When replicas_per_node_group is set, replica_count + 1 must equal num_node_groups * (replicas_per_node_group + 1)."
+    }
   }
 }
 
@@ -176,8 +188,11 @@ locals {
   instance_size             = split(".", var.node_type)[2]
   instances                 = sort(aws_elasticache_replication_group.this.member_clusters)
   owned_security_group_ids  = module.server_security_group.*.id
+  num_node_groups           = local.use_num_node_groups ? var.num_node_groups : null
   replica_enabled           = var.replica_count > 0
+  replicas_per_node_group   = local.use_num_node_groups ? coalesce(var.replicas_per_node_group, (local.instance_count / local.num_node_groups) - 1) : null
   shared_security_group_ids = var.server_security_group_ids
+  use_num_node_groups       = local.replica_enabled && var.num_node_groups != null
 
   instance_size_thresholds = {
     micro = 128

elasticache-redis/replication-group/variables.tf

@@ -48,6 +48,22 @@ variable "node_type" {
   description = "Node type for the Elasticache instance"
 }
 
+variable "num_node_groups" {
+  type        = number
+  default     = null
+  description = "Number of node groups (shards) to create when sharding is enabled"
+
+  validation {
+    condition     = var.num_node_groups == null || var.replica_count > 0
+    error_message = "num_node_groups can only be set when replica_count is greater than 0."
+  }
+
+  validation {
+    condition     = var.num_node_groups == null || var.num_node_groups <= (var.replica_count + 1)
+    error_message = "num_node_groups cannot exceed the total instance count implied by replica_count + 1."
+  }
+}
+
 variable "parameter_group_name" {
   type        = string
   description = "Parameter group name for the Redis cluster"
@@ -66,6 +82,17 @@ variable "replica_count" {
   description = "Number of read-only replicas to add to the cluster"
 }
 
+variable "replicas_per_node_group" {
+  type        = number
+  default     = null
+  description = "Number of replicas to create in each node group when num_node_groups is set"
+
+  validation {
+    condition     = var.replicas_per_node_group == null || var.num_node_groups != null
+    error_message = "replicas_per_node_group can only be set when num_node_groups is also set."
+  }
+}
+
 variable "replication_group_id" {
   description = "Override the ID for the replication group"
   type        = string