diff --git a/.claude/plans/percona-server.md b/.claude/plans/percona-server.md
new file mode 100644
index 0000000..6422f1e
--- /dev/null
+++ b/.claude/plans/percona-server.md
@@ -0,0 +1,249 @@
+# terraform-aws-percona-server
+
+Terraform module for Percona Server replica set with GTID replication, Orchestrator HA, and automated failover.
+
+GitHub Project: https://github.com/orgs/infrahouse/projects/7
+
+---
+
+## Requirements Summary
+
+### Infrastructure
+- Single ASG, odd number of instances (minimum 3)
+- ASG health check type: ELB (uses NLB target group health)
+- NLB with separate write/read target groups
+- DynamoDB table for locks and topology
+- S3 bucket for backups and binlogs
+
+### Database
+- Percona Server with GTID replication
+- One master, N replicas
+
+### Orchestrator
+- Sidecar on each node
+- Raft cluster with SQLite backend
+- Post-failover hooks:
+    - Update NLB target groups
+    - Update scale-in protection
+    - Update DynamoDB topology record
+
+### Provisioning
+- Puppet configures Percona and Orchestrator
+- Puppet registers instances with appropriate NLB target group
+- Custom facts provide cluster_id, dynamodb_table, s3_bucket
+
+### Bootstrap (Master Election)
+- Instance acquires DynamoDB lock (`lock-{cluster_id}`)
+- Reads topology record (`topology-{cluster_id}`)
+- If no master: become master, write topology, take full backup, release lock
+- If master exists: restore from backup, configure as replica, release lock
+- Uses existing DynamoDB lock class: https://github.com/infrahouse/infrahouse-core/blob/main/src/infrahouse_core/aws/dynamodb.py#L15
+
+### Backups
+- Tool: XtraBackup
+- Schedule: Weekly full + daily incremental, 00:{random_minute} UTC
+- Source: One replica (lock-based election via `backup-lock-{cluster_id}`)
+- Destination: S3
+- Retention: Configurable, default 4 weeks
+- Instance size: Up to 5TB (expect 4-6 hours for full backup/restore)
+
+### Binlog Archival
+- Real-time streaming via mysqlbinlog
+- One replica holds `binlog-lock-{cluster_id}`
+- Position tracked in DynamoDB (`binlog-position-{cluster_id}`)
+- GTID-based tracking for minimal data loss
+- Synced to S3 continuously
+
+### Health & Replacement
+- NLB health check detects dead Percona (port 3306)
+- ASG uses NLB health status to trigger instance replacement
+- Puppet on new instance handles re-registration
+
+### Target Group Registration
+
+| Event | Who Handles | Action |
+|-------|-------------|--------|
+| Initial master boot | Puppet | Register to write TG |
+| Initial replica boot | Puppet | Register to read TG |
+| Master dies (failover) | Orchestrator hook | Deregister old master from write TG, register new master to write TG |
+| Replica dies (replacement) | Puppet on new instance | Register to read TG |
+
+### DynamoDB Keys
+
+| Key | Purpose |
+|-----|---------|
+| `lock-{cluster_id}` | Master election lock |
+| `topology-{cluster_id}` | Master info (instance_id, private_ip) |
+| `backup-lock-{cluster_id}` | Backup leader election |
+| `binlog-lock-{cluster_id}` | Binlog streaming leader election |
+| `binlog-position-{cluster_id}` | Last synced GTID position |
+
+---
+
+## Project Items (Todo)
+
+### Infrastructure
+1. ASG configuration - single ASG, odd number validation, ELB health check type
+2. NLB with separate write/read target groups
+3. DynamoDB table for locks and topology
+4. S3 bucket for backups and binlogs
+
+### Percona Server
+5. Install Percona repository via Puppet
+6. Configure Percona Server with GTID replication
+7. Master election bootstrap logic with DynamoDB lock
+8. Scale-in protection for master instance
+
+### Orchestrator
+9. Orchestrator sidecar deployment (Raft + SQLite)
+10. Post-failover hook: Update NLB target groups
+11. Post-failover hook: Update scale-in protection
+12. Post-failover hook: Update DynamoDB topology record
+
+### Backups
+13. XtraBackup installation and configuration
+14. Weekly full backup with lock-based leader election
+15. Daily incremental backup
+16. S3 backup storage with configurable retention (default 4 weeks)
+17. Bootstrap backup from master (held until complete)
+
+### Binlog Archival
+18. Real-time binlog streaming setup (mysqlbinlog)
+19. Binlog streaming leader election with lock
+20. GTID position tracking in DynamoDB
+21. S3 binlog sync
+
+### Puppet Integration
+22. Custom facts for cluster_id, dynamodb_table, s3_bucket
+23. NLB target group registration logic
+24. Percona Server configuration management
+25. Orchestrator configuration management
+
+---
+
+## Module Inputs (Draft)
+
+```hcl
+variable "cluster_id" {
+  description = "Unique identifier for the Percona cluster"
+  type        = string
+}
+
+variable "instance_count" {
+  description = "Number of instances (must be odd, minimum 3)"
+  type        = number
+  default     = 3
+
+  validation {
+    condition     = var.instance_count >= 3 && var.instance_count % 2 == 1
+    error_message = "Instance count must be an odd number >= 3"
+  }
+}
+
+variable "instance_type" {
+  description = "EC2 instance type"
+  type        = string
+}
+
+variable "vpc_id" {
+  description = "VPC ID"
+  type        = string
+}
+
+variable "subnet_ids" {
+  description = "Subnet IDs for the ASG"
+  type        = list(string)
+}
+
+variable "backup_retention_weeks" {
+  description = "Number of weeks to retain backups"
+  type        = number
+  default     = 4
+}
+
+variable "percona_version" {
+  description = "Percona Server version (e.g., ps80, ps57)"
+  type        = string
+  default     = "ps80"
+}
+```
+
+---
+
+## Architecture Diagram
+
+```
+┌─────────────────────────────────────────────────────────┐
+│                     Single ASG                          │
+│  ┌─────────────┐  ┌─────────────┐  ┌─────────────┐     │
+│  │   Percona   │  │   Percona   │  │   Percona   │     │
+│  │   Master    │  │   Replica   │  │   Replica   │     │
+│  │ (protected) │  │             │  │             │     │
+│  │ Orchestrator│  │ Orchestrator│  │ Orchestrator│     │
+│  └─────────────┘  └─────────────┘  └─────────────┘     │
+└─────────────────────────────────────────────────────────┘
+         │                                    │
+         │ write TG                          │ read TG
+         ▼                                    ▼
+┌─────────────────────────────────────────────────────────┐
+│                         NLB                             │
+│    db-write.internal          db-read.internal          │
+└─────────────────────────────────────────────────────────┘
+
+┌─────────────────────────────────────────────────────────┐
+│                      DynamoDB                           │
+│  lock-{cluster_id}         topology-{cluster_id}        │
+│  backup-lock-{cluster_id}  binlog-lock-{cluster_id}     │
+│  binlog-position-{cluster_id}                           │
+└─────────────────────────────────────────────────────────┘
+
+┌─────────────────────────────────────────────────────────┐
+│                         S3                              │
+│  s3://{bucket}/{cluster_id}/full/{timestamp}/           │
+│  s3://{bucket}/{cluster_id}/incremental/{timestamp}/    │
+│  s3://{bucket}/{cluster_id}/binlogs/                    │
+└─────────────────────────────────────────────────────────┘
+```
+
+---
+
+## Failover Flow
+
+1. Master dies
+2. NLB health check fails, stops sending traffic to master
+3. ASG marks instance unhealthy (ELB health check type)
+4. Orchestrator (on replicas) detects master failure
+5. Orchestrator raft consensus elects new master
+6. Orchestrator post-failover hook:
+    - Deregisters old master from write TG
+    - Registers new master to write TG
+    - Removes scale-in protection from old master
+    - Adds scale-in protection to new master
+    - Updates DynamoDB topology record
+7. ASG terminates old master, launches replacement
+8. New instance boots, runs Puppet
+9. Puppet reads topology from DynamoDB, configures as replica
+10. Puppet registers with read TG
+
+---
+
+## Bootstrap Flow (First Cluster Creation)
+
+1. ASG launches 3 instances simultaneously
+2. Each instance runs Puppet
+3. Each tries to acquire `lock-{cluster_id}`
+4. Winner:
+    - Reads `topology-{cluster_id}` - not found
+    - Configures self as master
+    - Writes topology record (instance_id, private_ip)
+    - Takes full XtraBackup to S3
+    - Releases lock
+    - Registers with write TG
+    - Sets scale-in protection on self
+5. Losers (one at a time):
+    - Acquire lock
+    - Read topology - found master
+    - Restore from S3 backup
+    - Configure as replica with `MASTER_AUTO_POSITION=1`
+    - Release lock
+    - Register with read TG
diff --git a/repos.tf b/repos.tf
index 4c8517b..4cbc53e 100755
--- a/repos.tf
+++ b/repos.tf
@@ -217,6 +217,13 @@ locals {
         OPENVPN_CLIENT_SECRET : module.openvpn-oauth-client-id.secret_value
       }
     }
+    "terraform-aws-percona-server" = {
+      "description" = <<-EOT
+        Terraform module for Percona Server replica set with GTID replication,
+        Orchestrator HA, and automated failover.
+      EOT
+      "type"        = "terraform_module"
+    }
     "terraform-aws-pmm-ecs" = {
       "description" = <<-EOT
         Terraform module for deploying Percona Monitoring and Management (PMM) server