SKILL.md
AWS Troubleshooting Expertise
Investigation Methodology
- Identify the AWS resource/service involved
- Check resource status using describe functions
- Review CloudWatch logs for errors
- Check CloudWatch metrics for anomalies
- Analyze configuration for misconfigurations
- Synthesize and recommend
CloudWatch Logs Strategy
Partition First (CRITICAL)
Never dump all logs. Use aggregation queries first:
# Error rate over time
filter @message like /ERROR/
| stats count(*) as errors by bin(5m)
# Top error messages
filter @message like /Exception/
| stats count(*) by @message
| sort count desc
| limit 10
# Latency percentiles
stats pct(@duration, 50) as p50, pct(@duration, 99) as p99 by bin(5m)
# Unique error types
filter @message like /ERROR/
| parse @message /(?<error_type>[\w.]+Exception)/
| stats count(*) by error_typeQuery Flow
- Statistics first: Get error counts, distributions
- Identify time windows: Find when errors spiked
- Sample from spikes: Get specific examples
- Compare to baseline: Query same period yesterday/last week
Service-Specific Patterns
EC2 Issues
| Symptom | First Check | Typical Cause |
|---|---|---|
| Unreachable | describeec2instance |
Security group, stopped, status check failed |
| Performance | getcloudwatchmetrics (CPUUtilization) |
CPU exhaustion, network saturation |
| Disk full | getcloudwatchmetrics (DiskSpaceUtilization) |
Logs, temp files |
Key CloudWatch metrics for EC2:
- CPUUtilization
- NetworkIn, NetworkOut
- DiskReadOps, DiskWriteOps
- StatusCheckFailed
Lambda Issues
| Symptom | First Check | Typical Cause |
|---|---|---|
| Timeout | CloudWatch logs | External call slow, cold start, insufficient memory |
| Permission denied | CloudWatch logs | IAM role missing permissions |
| Memory error | CloudWatch metrics | Memory allocation too low |
| Cold starts | CloudWatch logs + metrics | Provisioned concurrency needed |
Key CloudWatch metrics for Lambda:
- Invocations
- Duration
- Errors
- Throttles
- ConcurrentExecutions
CloudWatch Insights for Lambda:
# Cold start analysis
filter @type = "REPORT"
| stats avg(@initDuration) as avg_cold_start,
count(@initDuration) as cold_starts,
count(*) as total_invocations
by bin(5m)
# Timeout analysis
filter @message like /Task timed out/
| stats count(*) by bin(5m)ECS/Fargate Issues
| Symptom | First Check | Typical Cause |
|---|---|---|
| Task failed | listecstasks |
Container crash, resource limits, image pull |
| Service unhealthy | listecstasks |
Health check failing, target group issues |
| Slow scaling | CloudWatch metrics | Insufficient capacity, service limits |
Investigation flow:
listecstasks- See task status and health- Check stopped reason in task description
- Review CloudWatch logs for the task
- Check container insights metrics
RDS Issues
| Symptom | First Check | Typical Cause |
|---|---|---|
| Connection refused | getrdsinstance_status |
Security group, stopped, maintenance |
| Slow queries | CloudWatch metrics | CPU, IOPS, connections |
| Storage full | CloudWatch metrics | Data growth, logs, snapshots |
Key CloudWatch metrics for RDS:
- CPUUtilization
- DatabaseConnections
- ReadIOPS, WriteIOPS
- FreeStorageSpace
- ReadLatency, WriteLatency
Common AWS Errors
Permission Errors
AccessDeniedException
UnauthorizedAccess→ Check IAM role/policy attached to the service
Throttling
Throttling
Rate exceeded
TooManyRequestsException→ Implement exponential backoff, request limit increase
Resource Not Found
ResourceNotFoundException
NoSuchEntity→ Verify resource name, region, account