RDS backups to S3

* Conditonally create a scheduled Fargate task to backup RDS databases
  to S3
* Fargate was chosen, so that it doesn't rely on the ECS infrastructure
  to run - meaning we can have RDS backups running even if an ECS
  Infrastructure hasn't been launched
* A CodeBuild project is created, which clones, builds and pushes the
  dalmatian-sql-backup image to ECR, ready to be used for the backup
  scheduled task. The Dockerfile for this is stored at
  https://github.com/dxw/dalmatian-sql-backup. This CodeBuild project is
  triggered every 24 hours to keep up with any updates.
* The backup task will backup all databases (except system databases),
  and store the SQL dumps into an S3 bucket, which can be configured
  with the retention time in days.

buildspecs/dalmatian-sql-backup.yml

@@ -0,0 +1,24 @@
+version: 0.2
+
+phases:
+  pre_build:
+    commands:
+      - echo "Build started on $(date)"
+      - echo "Logging in to Amazon ECR..."
+      - aws ecr get-login-password --region "$AWS_DEFAULT_REGION" | docker login --username AWS --password-stdin "$AWS_ACCOUNT_ID.dkr.ecr.$AWS_DEFAULT_REGION.amazonaws.com"
+      - |
+        if [ -n "$DOCKERHUB_USERNAME" ] && [ -n "DOCKERHUB_TOKEN" ];
+        then
+          echo "Logging into Dockerhub ...";
+          docker login --username "$DOCKERHUB_USERNAME" --password "$DOCKERHUB_PASSWORD";
+        fi;
+      - echo Building dalmatian-sql-backup docker image ...
+      - docker build -t dalmatian-sql-backup:latest .
+  build:
+    commands:
+      - echo Adding ECR repo tag...
+      - docker tag dalmatian-sql-backup:latest "$REPOSITORY_URI:latest"
+  post_build:
+    commands:
+      - echo Pushing the Docker image...
+      - docker push "$REPOSITORY_URI:latest"

container-definitions/app.json.tpl

@@ -8,6 +8,9 @@
     "logConfiguration": {
       "logDriver": "awslogs",
       "options": {
+        %{ if awslogs_stream_prefix != "" }
+        "awslogs-stream-prefix": "${awslogs_stream_prefix}",
+        %{ endif }
         "awslogs-group": "${cloudwatch_log_group}",
         "awslogs-region": "${region}"
       }
@@ -30,10 +33,10 @@
     ],
     %{ endif }
     %{ if environment != "[]" }
-    "environment": "${environment},
+    "environment": ${environment},
     %{ endif }
     %{ if secrets != "[]" }
-    "secrets": "${secrets},
+    "secrets": ${secrets},
     %{ endif }
     %{ if environment_file_s3 != "" }
     "environmentFiles": [

ecs-cluster-infrastructure-service-scheduled-task.tf

@@ -39,8 +39,9 @@ resource "aws_ecs_task_definition" "infrastructure_ecs_cluster_service_scheduled
       linux_parameters = jsonencode({
         initProcessEnabled = false
       })
-      cloudwatch_log_group = each.value["enable_cloudwatch_logs"] == true ? aws_cloudwatch_log_group.infrastructure_ecs_cluster_service[each.value["container_name"]].name : ""
-      region               = local.aws_region
+      cloudwatch_log_group  = each.value["enable_cloudwatch_logs"] == true ? aws_cloudwatch_log_group.infrastructure_ecs_cluster_service[each.value["container_name"]].name : ""
+      awslogs_stream_prefix = ""
+      region                = local.aws_region
     }
   )
   execution_role_arn       = aws_iam_role.infrastructure_ecs_cluster_service_task_execution[each.value["container_name"]].arn
@@ -123,7 +124,7 @@ resource "aws_iam_policy" "infrastructure_ecs_cluster_service_scheduled_task_pas
     "${path.root}/policies/pass-role.json.tpl",
     {
       role_arn = aws_iam_role.infrastructure_ecs_cluster_service_task_execution[each.value["container_name"]].arn
-      service  = "events.amazonaws.com"
+      service  = "ecs-tasks.amazonaws.com"
     }
   )
 }

ecs-cluster-infrastructure-service.tf

@@ -241,8 +241,9 @@ resource "aws_ecs_task_definition" "infrastructure_ecs_cluster_service" {
       linux_parameters = each.value["enable_execute_command"] == true ? jsonencode({
         initProcessEnabled = true
       }) : "{}"
-      cloudwatch_log_group = each.value["enable_cloudwatch_logs"] == true ? aws_cloudwatch_log_group.infrastructure_ecs_cluster_service[each.key].name : ""
-      region               = local.aws_region
+      cloudwatch_log_group  = each.value["enable_cloudwatch_logs"] == true ? aws_cloudwatch_log_group.infrastructure_ecs_cluster_service[each.key].name : ""
+      awslogs_stream_prefix = ""
+      region                = local.aws_region
     }
   )
   execution_role_arn       = aws_iam_role.infrastructure_ecs_cluster_service_task_execution[each.key].arn

ecs-entrypoints/rds-s3-backups-mysql.txt.tpl

@@ -0,0 +1 @@
+/usr/local/bin/ump-to-s3-mysql.sh -b ${s3_bucket_name}

ecs-entrypoints/rds-s3-backups-postgres.txt.tpl

@@ -0,0 +1 @@
+/usr/local/bin/dump-to-s3-postgres.sh -b ${s3_bucket_name}

kms-infrastructure.tf

@@ -46,6 +46,18 @@ resource "aws_kms_key" "infrastructure" {
           for k, v in local.infrastructure_ecs_cluster_services : aws_iam_role.infrastructure_ecs_cluster_service_task_execution[k].arn if v["enable_execute_command"] == true && local.infrastructure_ecs_cluster_enable_execute_command_logging
         ])
       }
+      )}${length(local.infrastructure_rds) > 0 && local.infrastructure_kms_encryption && local.enable_infrastructure_rds_backup_to_s3 ? "," : ""}
+      ${templatefile("${path.root}/policies/kms-key-policy-statements/role-allow-encrypt.json.tpl",
+      {
+        role_arns = jsonencode([
+          for k, v in local.infrastructure_rds : aws_iam_role.infrastructure_rds_s3_backups_task[k].arn if local.enable_infrastructure_rds_backup_to_s3
+        ])
+      }
+      )}${length(local.infrastructure_rds) > 0 && local.infrastructure_kms_encryption && local.enable_infrastructure_rds_backup_to_s3 ? "," : ""}
+      ${templatefile("${path.root}/policies/kms-key-policy-statements/cloudwatch-logs-allow.json.tpl",
+      {
+        log_group_arn = length(local.infrastructure_rds) > 0 && local.infrastructure_kms_encryption && local.enable_infrastructure_rds_backup_to_s3 ? "arn:aws:logs:${local.aws_region}:${local.aws_account_id}:log-group:${local.resource_prefix}-infrastructure-rds-s3-backups-*" : ""
+      }
       )}${contains([for k, v in local.custom_s3_buckets : (v["cloudfront_dedicated_distribution"] == true || v["cloudfront_infrastructure_ecs_cluster_service"] != null) && v["create_dedicated_kms_key"] == false ? true : false], true) && local.infrastructure_kms_encryption ? "," : ""}
       ${templatefile("${path.root}/policies/kms-key-policy-statements/cloudfront-distribution-allow.json.tpl",
       {

local-exec-scripts/trigger-codedeploy-project.sh

@@ -0,0 +1,70 @@
+#!/bin/bash
+
+# exit on failures
+set -e
+set -o pipefail
+
+usage() {
+  echo "Usage: $(basename "$0") [OPTIONS]" 1>&2
+  echo "  -h   - help"
+  echo "  -n   - CodeBuild project name"
+  exit 1
+}
+
+while getopts "n:h" opt; do
+  case $opt in
+    n)
+      PROJECT_NAME=$OPTARG
+      ;;
+    h)
+      usage
+      ;;
+    *)
+      usage
+      ;;
+  esac
+done
+
+if [ -z "$PROJECT_NAME" ]
+then
+  usage
+fi
+
+BUILD_ID="$(
+  aws codebuild start-build \
+  --project-name "$PROJECT_NAME" \
+  | jq -r '.build.id'
+)"
+
+echo "Triggered $PROJECT_NAME CodeBuild project ($BUILD_ID) ..."
+
+COMPLETED=""
+while [ -z "$COMPLETED" ]
+do
+  sleep 10
+  echo "Checking progress of CodeBuild  $BUILD_ID ..."
+  BUILD_PROGRESS="$(
+    aws codebuild batch-get-builds \
+    --ids "$BUILD_ID" \
+  )"
+  COMPLETED="$(
+    echo "$BUILD_PROGRESS" \
+    | jq -r \
+    '.builds[0].phases[] | select(.phaseType == "COMPLETED")'
+  )"
+done
+echo "CodeBuild $BUILD_ID Completed, checking for failures ..."
+
+FAILURES="$(
+  echo "$BUILD_PROGRESS" \
+  | jq -r \
+  '.builds[0].phases[] | select(.phaseStatus == "FAILED")'
+)"
+
+if [ -n "$FAILURES" ]
+then
+  echo "$FAILURES"
+  exit 1
+fi
+
+echo "CodeBuild $BUILD_ID completed without failures"

locals.tf

@@ -30,12 +30,14 @@ locals {
     length(local.custom_s3_buckets) != 0 ||
     local.enable_cloudformatian_s3_template_store ||
     local.enable_infrastructure_vpc_transfer_s3_bucket ||
-    local.infrastructure_ecs_cluster_enable_execute_command_logging
+    local.infrastructure_ecs_cluster_enable_execute_command_logging ||
+    local.enable_infrastructure_rds_backup_to_s3
   )
   logs_bucket_s3_source_arns = concat(
     length(local.infrastructure_ecs_cluster_services) != 0 ? [aws_s3_bucket.infrastructure_ecs_cluster_service_build_pipeline_artifact_store[0].arn] : [],
     local.enable_infrastructure_vpc_transfer_s3_bucket ? [aws_s3_bucket.infrastructure_vpc_transfer[0].arn] : [],
-    [for k, v in local.custom_s3_buckets : aws_s3_bucket.custom[k].arn]
+    [for k, v in local.custom_s3_buckets : aws_s3_bucket.custom[k].arn],
+    local.enable_infrastructure_rds_backup_to_s3 ? [aws_s3_bucket.infrastructure_rds_s3_backups[0].arn] : [],
   )
   logs_bucket_logs_source_arns = concat(
     local.infrastructure_vpc_flow_logs_s3_with_athena ? ["arn:aws:logs:${local.aws_region}:${local.aws_account_id}:*"] : []
@@ -232,6 +234,17 @@ locals {
     "mysql"    = 3306
     "postgres" = 5432
   }
+  rds_s3_backups_container_entrypoint_file = {
+    "mysql"    = "${path.root}/ecs-entrypoints/rds-s3-backups-mysql.txt.tpl"
+    "postgres" = "${path.root}/ecs-entrypoints/rds-s3-backups-postgres.txt.tpl"
+  }
+  enable_infrastructure_rds_backup_to_s3          = var.enable_infrastructure_rds_backup_to_s3
+  infrastructure_rds_backup_to_s3_cron_expression = var.infrastructure_rds_backup_to_s3_cron_expression
+  infrastructure_rds_backup_to_s3_retention       = var.infrastructure_rds_backup_to_s3_retention
+  enable_infrastructure_rds_tooling_ecs_cluster = anytrue([
+    local.enable_infrastructure_rds_backup_to_s3,
+  ])
+  infrastructure_rds_tooling_ecs_cluster_name = "${local.resource_prefix}-infrastructure-rds-tooling"
 
   infrastructure_elasticache_defaults = var.infrastructure_elasticache_defaults
   infrastructure_elasticache_keys     = length(var.infrastructure_elasticache) > 0 ? keys(values(var.infrastructure_elasticache)[0]) : []

policies/codebuild-allow-builds.json.tpl

@@ -0,0 +1,16 @@
+{
+  "Version": "2012-10-17",
+  "Statement": [
+    {
+      "Action": [
+        "codebuild:StartBuild",
+        "codebuild:StopBuild",
+        "codebuild:BatchGet*",
+        "codebuild:Get*",
+        "codebuild:List*"
+      ],
+      "Effect": "Allow",
+      "Resource": "*"
+    }
+  ]
+}

policies/secrets-manager-get-secret-value.json.tpl

@@ -0,0 +1,12 @@
+{
+  "Version": "2012-10-17",
+  "Statement": [
+    {
+      "Effect": "Allow",
+      "Action": [
+        "secretsmanager:GetSecretValue"
+      ],
+      "Resource": ${secret_name_arns}
+    }
+  ]
+}

rds-infrastructure-cloudwatch-logs.tf

@@ -10,3 +10,12 @@ resource "aws_cloudwatch_log_group" "infrastructure_rds_exports" {
   retention_in_days = 30
   kms_key_id        = local.infrastructure_kms_encryption ? aws_kms_key.infrastructure[0].id : null
 }
+
+resource "aws_cloudwatch_log_group" "infrastructure_rds_s3_backups" {
+  for_each = local.enable_infrastructure_rds_backup_to_s3 ? local.infrastructure_rds : {}
+
+  name              = "${local.resource_prefix}-infrastructure-rds-s3-backups-${each.key}"
+  retention_in_days = local.infrastructure_rds_backup_to_s3_retention
+  kms_key_id        = local.infrastructure_kms_encryption ? aws_kms_key.infrastructure[0].arn : null
+  skip_destroy      = true
+}

rds-infrastructure-s3-backups-ecr.tf

@@ -0,0 +1,17 @@
+resource "aws_ecr_repository" "infrastructure_rds_s3_backups" {
+  count = local.enable_infrastructure_rds_backup_to_s3 ? 1 : 0
+
+  name = "${local.resource_prefix}-rds-s3-backups"
+
+  #tfsec:ignore:aws-ecr-enforce-immutable-repository
+  image_tag_mutability = "MUTABLE"
+
+  encryption_configuration {
+    encryption_type = local.infrastructure_kms_encryption ? "KMS" : "AES256"
+    kms_key         = local.infrastructure_kms_encryption ? aws_kms_key.infrastructure[0].arn : null
+  }
+
+  image_scanning_configuration {
+    scan_on_push = true
+  }
+}