AWS Migration Hub Implementation Guide - AI-Optimized Technical Reference

Critical Configuration Requirements

Discovery Agent Installation

• CPU Impact: 40%+ on legacy servers (older than 5 years)
• Memory Consumption: 500MB+ after 2-3 weeks (memory leak)
• Performance Threshold: Systems with <4GB RAM will experience degradation
• Failure Scenario: Windows 2008 R2 domain controllers become unresponsive for 20+ minutes during registry scanning

Production-Ready Configuration:

# /opt/aws/discovery/config/agent.properties
collection_interval=3600  # Change from default 900 seconds

Required Maintenance:

# Weekly restart to prevent memory leak crashes
0 2 * * 0 systemctl restart aws-discovery-daemon

Network Discovery Limitations

• Coverage Gap: Misses 20% of critical dependencies
• Sampling Issue: Only captures active connections during scan period
• Hidden Dependencies: Monthly/quarterly jobs, backup processes, scheduled tasks
• Minimum Discovery Period: 14 days to catch weekly/monthly processes

Authentication Requirements

Actual Permissions Needed (beyond documented):

• discovery:*
• mgh:*
• AWSApplicationMigrationAgentPolicy
• AWSApplicationMigrationReplicationServerPolicy
• Custom CloudWatch logs access policy

Debug Process: Use CloudTrail to identify failing API calls, then add specific permissions

Home Region Constraints

• Critical Limitation: Cannot change home region without AWS Support intervention
• Resolution Time: 1-2 business days via AWS Support
• Impact: All migration data locked to initial region selection

Resource Requirements and Time Estimates

Migration Timeline Reality

Server Size	Network Speed	Actual Time
500GB	100Mbps	6-12 hours
2TB with database	100Mbps	2-3 days
10TB file server	100Mbps	1-2 weeks

Rule of thumb: Add 50% buffer for network issues and AWS throttling

API Rate Limiting

• Threshold: 100+ servers trigger undocumented rate limits
• Symptoms: HTTP 429 errors, console unresponsiveness
• Workaround: Exponential backoff, batch API calls
• Monitoring: CloudWatch metrics for API error spikes

Critical Failure Modes

Discovery Agent Failures

1. CPU Overload on Legacy Systems

   • Cause: Agent scans every process/connection every 15 minutes
   • Impact: Single-core systems become unresponsive
   • Fix: Increase scan interval to 1 hour

2. Memory Leak Crashes

   • Cause: Known leak in agent process
   • Timeline: 2-3 weeks to failure
   • Detection: Process >500MB RAM usage
   • Solution: Weekly automated restarts

3. Network Connectivity Issues

   • Symptom: Agent shows "healthy" but no data in console
   • Cause: Firewall/proxy blocking AWS endpoints
   • Test: curl -I https://application-discovery.us-west-2.amazonaws.com
   • Fix: Configure proxy or open ports 443, 8888

Migration Tracking Failures

1. Status Mapping Breaks

   • Frequency: 40% of migrations lose tracking
   • Cause: Automatic mapping fails between tools and discovered servers
   • Manual Fix: Migration Hub → Updates → Edit → Manually map servers

2. Application Group Logic Fails

   • Issue: Servers assigned to wrong applications or multiple groups
   • Impact: "Partially migrated" status for completed migrations
   • Solution: Delete auto-generated groups, create manual groups

3. Multi-Region Tracking Impossible

   • Limitation: Single home region view only
   • Impact: Cannot track migrations spanning multiple regions
   • Workaround: Custom CloudWatch dashboards and Lambda functions

Migration Execution Failures

1. Boot Failures After Migration

   • Frequency: 30% of Windows servers, 15% of custom Linux
   • Cause: Hardware differences, driver issues, boot sector problems
   • Emergency Fix:
     • Launch rescue instance
     • Mount migrated volume as secondary disk
     • Fix /etc/fstab or Windows registry for new hardware

2. Application Non-Functionality

   • Issue: Server migrates successfully but application fails
   • Common Causes: Database connection strings, license servers, network routing
   • Testing Required: 2-4 weeks post-migration troubleshooting

Performance Monitoring Inadequacies

Discovery Agent Metrics Limitations

• Averaging Period: 15-minute intervals mask peak loads
• Missing Data: Night-time backups, month-end processing spikes
• Impact: Undersized AWS instances cause performance issues

Better Monitoring Approach:

• Maintain existing monitoring tools during discovery
• Export VMware vCenter performance data
• Use AWS Systems Manager for detailed metrics
• Run stress tests for actual resource requirements

Decision Support Matrix

Agent vs Agentless Discovery

Criteria	Agent-Based	Agentless
Dependency Mapping	Complete (with caveats)	Basic specs only
Performance Impact	High (40% CPU)	None
Server Compatibility	All platforms	VMware vCenter 5.5+ only
Application Visibility	Process-level	None
Installation Complexity	High	Medium

When to Use Alternative Tools

• CloudEndure/AWS MGN: Better for large-scale migrations
• Turbonomic: More accurate right-sizing analysis
• Zerto: Enterprise-grade replication with better tracking

Emergency Procedures

Agent Removal (When Breaking Server)

sudo systemctl stop aws-discovery-daemon
sudo systemctl disable aws-discovery-daemon
sudo /opt/aws/discovery/uninstall
sudo rm -rf /opt/aws/discovery
# If uninstaller fails:
sudo pkill -f discovery

Migration Rollback Triggers

• Boot failures exceeding downtime window
• Database corruption or data inconsistency
• Unusable application performance
• Network connectivity preventing user access

Support Escalation Paths

1. Basic Issues: AWS re:Post Migration Hub questions
2. Critical Failures: AWS Enterprise Support (24/7)
3. Complex Migrations: AWS Professional Services
4. Large Scale: AWS Migration Acceleration Program

Operational Intelligence

What AWS Documentation Doesn't Tell You

• "Minimal performance impact" = 40% CPU on legacy systems
• "Automated mapping" works 60% of the time
• Rate limits exist but aren't documented
• Memory leaks require weekly restarts
• Boot failure rate is 30% for Windows migrations

Hidden Costs

• Time Investment: 2-4 weeks post-migration troubleshooting per application
• Expertise Required: Network engineering, systems administration, AWS architecture
• Infrastructure: Separate testing environment mandatory
• Support: Enterprise Support recommended for production migrations

Success Criteria

• Technical: Application functionality, not just server replication
• Performance: Stress testing under production loads
• Integration: All dependencies and scheduled jobs working
• Monitoring: Full observability stack operational
• Rollback: Tested and documented fallback procedures

Breaking Points

• 100+ servers: API rate limiting kicks in
• Legacy servers (>5 years): Performance degradation likely
• Complex applications: Manual dependency mapping required
• Multi-region: Native tracking fails, custom solution needed
• Large datasets (>2TB): Week+ migration windows required

This guide prioritizes real-world operational intelligence over vendor marketing claims, focusing on what actually breaks and how to fix it.