database-migration-service-expertise

Files (13)

• .skilleval.yaml
• CHANGELOG.md
• README.md
• SKILL.md
• benchmark.json · evals/
• eval_queries.json · evals/
• evals.json · evals/
• dms-context.json · evals/files/
• report.json · evals/
• trigger_report.json · evals/
• dms-best-practices.md · references/
• dms-validation-checklist.md · references/
• dms-version-reference.md · references/

SKILL.md (first lines)

# AWS DMS Operational Review & Troubleshooting

Conduct comprehensive operational reviews, best practices validation, troubleshooting,
and cost optimization assessments for AWS Database Migration Service (DMS) environments
aligned with AWS Well-Architected principles.

## When to Use

Activate this skill when the user asks to:
- Review, audit, or assess DMS configurations
- Check DMS best practices compliance
- Troubleshoot DMS task failures, latency, or connectivity issues
- Evaluate DMS cost optimization opportunities
- Plan or execute a database migration cutover
- Assess DMS version deprecations and upgrade paths
- Run a pre-migration assessment
- Evaluate DMS Serverless vs provisioned suitability
- Investigate data validation mismatches
- Review DMS homogeneous data migration options

## Step 1: Identify Target Resources

Ask the user which DMS resources to review. Accept:
- Specific replication instance names and regions
- Specific replication task identifiers
- "all instances" in a specific region
- "all instances in all regions"

Query AWS APIs to discover DMS resources:
- List replication instances across the configured account regions
- List replication tasks and their states
- List endpoints and connectivity status

## Step 2: Collect DMS Configuration

For EACH replication instance and its associated tasks, gather the following data using AWS APIs.

### 2.1 Replication Instance Config
Engine version, instance class, Multi-AZ, storage, public accessibility, KMS encryption, auto-upgrade, VPC/subnet configuration, pending maintenance actions.

```bash
aws dms describe-replication-instances --profile <profile> --region <region>
aws dms describe-pending-maintenance-actions --profile <profile> --region <region>
```

### 2.2 Replication Tasks
Task identifiers, status, migration type, table mappings, task settings, CDC start positions, last failure messages, stop reasons, replication task statistics.

```bash
aws dms describe-replication-tasks --profile <profile> --region <region>
```

### 2.3 Endpoints
Endpoint identifiers, types (source/target), engine names, SSL mode, connection status, Secrets Manager integration, security group configurations.

```bash
aws dms describe-endpoints --profile <profile> --region <region>
aws dms describe-connections --filter Name=endpoint-arn,Values=<endpoint-arn> --profile <profile> --region <region>
```

### 2.4 Table Statistics
Per-table state, row counts, validation state, error rows, DDL counts.

```bash
aws dms describe-table-statistics --replication-task-arn <task-arn> --profile <profile> --region <region>
```

### 2.5 DMS Serverless (if applicable)
Replication configs, serverless replications, DCU utilization.

```bash
aws dms describe-replication-configs --profile <profile> --region <region>
aws dms describe-replications --profile <profile> --region <region>
```

### 2.6 Version and Deprecation Data
Available engine versions, pending maintenance actions, AWS Health events.

```bash
aws dms describe-orderable-replication-instances --query "OrderableReplicationInstances[*].EngineVersion" --profile <profile> --region <region>
aws health describe-events --filter '{"services":["DMS"],"eventTypeCategories":["scheduledChange"]}' --profile <profile> --region us-east-1
```

> **Note:** AWS Health API is a global service and must be called with `--region us-east-1` regardless of where DMS resources are deployed.

## Step 3: Collect Observability Data (7-Day Historical)

### 3.1 CloudWatch Metrics (7 days)

**Replication Instance Metrics** (namespace: AWS/DMS):
- CPUUtilization (Average, Maximum)
- FreeableMemory (Average, Minimum)
- SwapUsage (Average, Maximum)
- DiskQueueDepth (Average, Maximum)
- ReadIOPS / WriteIOPS (Average, Maximum)
- ReadLatency / WriteLatency (Average, Maximum)
- NetworkTransmitThroughput (Average)
- NetworkReceiveThroughput (Average)

**Task Metrics** (namespace: AWS/DMS):
- CDCLatencySource (Average, Maximum)
- CDCLatencyTarget (Average, Maximum)
- CDCIncomingChanges (Average, Maximum)
- CDCChangesMemorySource (Average, Maximum)
- CDCChangesDiskSource (Maximum)
- CDCChangesDiskTarget (Maximum)
- FullLoadThroughputBandwidthTarget (Average)
- FullLoadThroughputRowsTarget (Average)

**DMS Serverless** (namespace: AWS/DMS, if applicable):
- ServerlessTotalCapacityUnits (Average, Maximum)

### 3.2 CloudWatch Logs (7 days)

Query task logs for error patterns:
- `ERROR` — general errors (count)
- Network/connectivity errors (`Network error`, `Connection lost`, `timeout`)
- Memory/resource errors (`out of memory`, `OOM`, `swap`)
- CDC-specific errors (WAL, binlog, LogMiner patterns)

From aws/tools-for-devops-agent, Apache-2.0.

More from aws/tools-for-devops-agent