AWS Maintenance

Monitoring

CloudWatch Logs

Docket outputs structured JSON logs. Send them to CloudWatch:

npm start | aws logs push --log-group docket --log-stream $(hostname)

CloudWatch Metrics

Key metrics to alarm on:

• Bedrock invocation errors
• DynamoDB throttled requests
• SQS queue depth
• S3 4xx/5xx errors

DynamoDB Metrics

Monitor in the DynamoDB console:

• Consumed Read/Write Capacity
• Throttled requests
• Table size

Scaling

DynamoDB

Switch to on-demand for unpredictable workloads:

aws dynamodb update-table \
  --table-name docket \
  --billing-mode PAY_PER_REQUEST

For predictable workloads, use provisioned capacity with auto-scaling:

aws dynamodb update-table \
  --table-name docket \
  --provisioned-throughput ReadCapacityUnits=100,WriteCapacityUnits=50

S3

S3 scales automatically. Monitor request rates and consider:

• S3 Transfer Acceleration for global users
• CloudFront in front of S3 for caching

SQS

For high throughput:

• Use batching: SendMessageBatch instead of SendMessage
• Enable long polling: ReceiveMessageWaitTimeSeconds=20
• Set up dead-letter queues for failed jobs

Cost Control

Service	Cost Driver
Bedrock	Input + output tokens per model
DynamoDB	Storage + read/write capacity
S3	Storage + requests + egress
SQS	Number of API requests

Set up AWS Budgets alerts at 80% of expected spend.

Backups

DynamoDB

Enable point-in-time recovery:

aws dynamodb update-continuous-backups \
  --table-name docket \
  --point-in-time-recovery-specification PointInTimeRecoveryEnabled=true

S3

Enable versioning:

aws s3api put-bucket-versioning \
  --bucket docket-blobs-your-account-id \
  --versioning-configuration Status=Enabled

Cross-Region Replication

For disaster recovery, replicate DynamoDB and S3 to a secondary region.