drift-detector（漂移探测器）

Infrastructure Drift Detector

Detect, track, and resolve infrastructure drift between Terraform state and actual cloud resources to maintain Infrastructure as Code integrity.

Core Principle: Your cloud should match your code.

Note: This skill uses snyk iac describe CLI command (requires shell execution).

Quick Start

# Basic drift scan against a local Terraform state file
snyk iac describe --from=tfstate://terraform.tfstate

# Output as JSON for further analysis
snyk iac describe --from=tfstate://terraform.tfstate --json > drift-report.json

Prerequisites

• Terraform project with state file (local or remote)
• Cloud provider credentials configured
• snyk CLI installed
• Network access to cloud APIs

Supported Cloud Providers

For a full list of supported resource types per provider, see SERVICES.md.

Phase 1: Setup

Goal: Configure drift detection environment.

Step 1.1: Verify Terraform State

Check for Terraform state:

Local state:

ls terraform.tfstate

Remote state (S3 backend):

terraform {
  backend "s3" {
    bucket = "my-terraform-state"
    key    = "state/terraform.tfstate"
    region = "us-east-1"
  }
}

Step 1.2: Verify Cloud Credentials

AWS:

aws sts get-caller-identity

Azure:

az account show

GCP:

gcloud auth application-default print-access-token

Phase 2: Run Drift Detection

Goal: Identify differences between IaC and actual cloud state.

Step 2.1: Basic Drift Scan

snyk iac describe --from=tfstate://terraform.tfstate

Step 2.2: Remote State Scan

For S3 backend:

snyk iac describe --from=tfstate+s3://my-bucket/state.tfstate

For Terraform Cloud:

snyk iac describe \
  --from=tfstate+tfcloud://organization/workspace \
  --tfc-token=$TFC_TOKEN

Step 2.3: Specific Service Scan

To focus on specific AWS services:

snyk iac describe \
  --from=tfstate://terraform.tfstate \
  --service=aws_s3,aws_ec2,aws_rds

Step 2.4: JSON Output for Analysis

snyk iac describe \
  --from=tfstate://terraform.tfstate \
  --json > drift-report.json

Phase 3: Analyze Results

Goal: Understand and categorize drift.

Step 3.1: Drift Categories

Step 3.2: Generate Report

## Infrastructure Drift Report

Scan Date: 2024-01-15
Terraform State: s3://my-bucket/prod.tfstate
Cloud Provider: AWS (us-east-1)

### Summary
- Unmanaged Resources: 12 (High)
- Changed Resources:    5  (Medium)
- Missing Resources:    2  (Low)
- Total Drift:         19

### Unmanaged Resources (Not in Terraform)
- aws_s3_bucket      | prod-logs-manual   | High     | Import or delete
- aws_security_group | sg-temp-access     | Critical | Review and remove

### Changed Resources (Modified Outside Terraform)
- aws_security_group.web | ingress: [443]   → ingress: [443, 22]  | High
- aws_rds_instance.main  | multi_az: true   → multi_az: false      | Critical

Step 3.3: Risk Assessment

Prioritize Critical issues first (e.g. SSH opened to 0.0.0.0/0, production HA disabled), then High risk issues (e.g. unmanaged IAM users or security groups). Document the affected resource, the risk, and the intended remediation action for each finding.

Phase 4: Remediation

Goal: Resolve drift and restore IaC integrity.

Step 4.1: Import Unmanaged Resources

For resources that should be in Terraform:

# Generate import block
terraform import aws_s3_bucket.manual_bucket prod-logs-manual

# Or use import block (Terraform 1.5+)

import {
  to = aws_s3_bucket.manual_bucket
  id = "prod-logs-manual"
}

Step 4.2: Remove Unauthorized Resources

For resources that shouldn't exist:

# After verification, delete unmanaged resources
aws s3 rb s3://unauthorized-bucket --force
aws ec2 terminate-instances --instance-ids i-temp-server

Step 4.3: Revert Changes

For resources modified outside Terraform:

# Re-apply Terraform to restore intended state
terraform apply

Step 4.4: Update Terraform (Adopt Changes)

If the manual change should be kept:

# Update Terraform to match new reality
resource "aws_security_group" "web" {
  # Add the new rule
  ingress {
    from_port   = 22
    to_port     = 22
    protocol    = "tcp"
    cidr_blocks = ["10.0.0.0/8"]  # Restrict if keeping
  }
}

Phase 5: Prevention

Goal: Prevent future drift.

Step 5.1: Generate Exclude Policy

For expected drift (auto-scaling, etc.):

snyk iac update-exclude-policy \
  --exclude-unmanaged \
  --exclude-changed

This creates a .snyk policy file:

exclude:
  iac-drift:
    - aws_autoscaling_group.*
    - aws_ecs_service.*:desiredCount

Step 5.2: CI/CD Integration

Add drift detection to CI/CD:

# GitHub Actions example
- name: Check for Infrastructure Drift
  run: |
    snyk iac describe \
      --from=tfstate+s3://my-bucket/prod.tfstate \
      --json > drift.json

    # Fail if unmanaged resources found
    if [ $(jq '.summary.total_unmanaged' drift.json) -gt 0 ]; then
      echo "Drift detected!"
      exit 1
    fi

Step 5.3: Regular Audits

Schedule regular drift audits:

Common Scenarios

For detailed worked examples, see EXAMPLES.md. Brief references:

• Post-Incident Audit: Run drift detection with JSON output, filter for security-related resources, identify unauthorized changes, generate incident report, remediate and document.
• Pre-Deployment Check: Run drift detection, fail deployment if drift exists, resolve drift first, then proceed with deployment.
• Shadow IT Discovery: Run drift detection, filter to unmanaged resources, categorize by owner/purpose, import or remove as appropriate.

Error Handling

State Access Error

Error: Could not read Terraform state

Solutions:
1. Verify state file path
2. Check S3/backend permissions
3. Ensure terraform init has been run

Cloud Credential Error

Error: Authentication failed

Solutions:
1. Verify cloud credentials
2. Check IAM permissions for describe/list
3. Ensure credentials not expired

Service Not Supported

Warning: Service X not supported

Solutions:
1. Check supported services list
2. Use Terraform plan comparison instead
3. Report to Snyk for feature request

Constraints

1. Read-only: This skill only detects drift, doesn't modify resources
2. Credentials required: Needs cloud provider access
3. Service coverage: Not all resource types supported
4. State required: Must have Terraform state to compare
5. Network required: Needs access to cloud APIs