Release Notes

Roadrunner SMB™ 1.0.5

Release date: June 2026
Availability: AWS Marketplace

Roadrunner SMB is an SMB3 file gateway for Amazon EFS® that deploys into your existing VPC via CloudFormation®. Windows® clients get domain-authenticated shares with native ACL semantics, while storage stays elastic on EFS. A browser-based Admin UI handles first-time setup, share management, and cluster health — with Network Load Balancer front-ends for SMB and HTTPS admin access.

Get started

VPC Prerequisites — confirm your VPC before launch
Quick Start Guide — subscribe, deploy, join Active Directory®, create your first share

Key features

Active Directory integration — domain join, Kerberos, and Windows ACL enforcement on EFS-backed shares
Configurable cluster — 1 to 4 ECS nodes; NLB load-balances SMB (TCP/445) across healthy nodes
Dual-zone and Single-AZ — HA production layouts (Dual-AZ) or Single-AZ for evaluation and non-critical workloads
Admin UI — HTTPS on port 443 via the stack NLB (AdminUIUrl); guided First-Time Setup wizard
Elastic SMB™ metadata acceleration — L1 cache and DynamoDB®-backed distributed state for Windows and VDI workloads
Automatic rolling software updates — new container and host releases are detected and scheduled automatically (see Software updates below)
AWS® Marketplace usage metering — billable dimensions: ShareHours and ManagedStorageGBHours
Customer-supplied EFS — you create EFS filesystems and access points; select them when creating shares in the Admin UI

AWS Marketplace billing — EFS storage

Roadrunner SMB charges for share-hours and share-mounted EFS storage in GB-hours:

ShareHours — one unit per enabled, billable share for each completed hour.
ManagedStorageGBHours — based on EFS filesystem size queried for each filesystem attached to billable shares.

Important: Metering uses the entire EFS filesystem size, not only the data under the share’s access-point path. If you mount an existing EFS filesystem with data already on it, you are billed for all GB in that filesystem. Plan capacity before attaching production filesystems.

If multiple shares use the same EFS filesystem, that filesystem is counted once per hour (not per share).

Roadrunner SMB does not create customer share filesystems or access points — create them in AWS and pick them during share creation. We recommend not changing EFS filesystem defaults unless your standards require it.

See Known Limitations for detail.

Software updates

Roadrunner SMB supports automatic rolling updates. When new software versions become available, they are automatically recognized and scheduled:

Container updates — delivered via Amazon ECS rolling task replacement (application image only; EC2 hosts unchanged).
EC2 host updates — delivered via the Auto Scaling group (ASG) when a newer host AMI or launch template is applied (for example Amazon Linux 2023 security or platform refreshes).

Cluster nodes run standard, unmodified Amazon Linux 2023 EC2 instance AMIs with Roadrunner SMB host utilities installed at bootstrap. ECS and ASG perform rolling replacements to preserve high availability when two or more cluster nodes are configured. Single-node deployments update with a brief service interruption.

Review release notes before production updates when your change-control process requires it. See Known Limitations — Software Updates for scheduling and operator controls.

Version 1.0.5 — cluster sizing and availability

Cluster size	Typical use	Availability posture
1 node	Evaluation, non-HA	Single point of failure
2 nodes (default)	Production HA minimum	99.9% availability
3–4 nodes	Production HA at scale	99.99% availability

Default: 2 ECS cluster nodes (Dual-AZ delivery option).
HA minimum: 2 nodes across Availability Zones with NLB health checks and coordinated cluster state.
SMB path: NLB distributes SMB (TCP/445) to healthy nodes from sources allowed by SmbClientCidr (VPC/private client networks by default).
Deployments: Dual-zone for production HA; Single-AZ for evaluation and non-critical use.

EC2 hosts and ECS tasks

Steady state: ECS desired task count and ASG desired capacity both equal ClusterSize — one Roadrunner SMB ECS task per EC2 host.
Rolling updates: ASG max size is ClusterSize + 1, allowing a temporary extra host while ECS replaces tasks or the ASG rolls instances. Capacity returns to ClusterSize when the update completes.
Not a permanent spare: under normal operation you run N EC2 instances for N ECS tasks, not N+1 standing capacity.

See Deployment Guide — Scaling for detail.

Stopping the cluster (scale to zero)

The Admin UI Cluster slider supports 1–4 only — not zero. To stop compute during evaluations or periods of inactivity:

Recommended: CloudFormation stack update with ApplianceState=Stopped (scales ECS and ASG to zero while retaining EFS and DynamoDB).
Manual: set the ASG desired capacity to 0, then terminate any remaining EC2 instances in the AWS Console.

To bring the cluster back online, update the CloudFormation stack with ApplianceState=Running and the desired Cluster Node Count, or set the ASG back to your target ClusterSize (align ApplianceState if you stopped manually). See FAQ — scale to zero.

Performance guidance

Roadrunner SMB is network-intensive (SMB front-end, EFS/NFS back-end). For maximum throughput, choose high-networking EC2 instance types at deploy time and ensure your VPC provides adequate NAT or VPC endpoint paths for AWS API traffic. File data is not cached — read/write performance follows EFS throughput and latency.

Known limitations and planning constraints

Read the complete list in Known Limitations. The items below are the most common pre-production surprises at GA.

Deployment and topology

One appliance per AWS Region per account — cluster ID is fixed (rrsmb-prod); a second stack in the same Region is not supported.
1–4 nodes only — sizes outside that range are not tested or supported.
1 node = evaluation and non-HA — no redundancy; brief SMB interruption during software updates.
Single-AZ — supported for lab and non-critical use; no AZ-level fault tolerance. Use Dual-AZ + 2+ nodes for production HA.
No cross-region DR — single-region deployment; plan recovery within the deployed Region.
Customer-supplied EFS — you create EFS filesystems and access points before creating shares; Roadrunner SMB does not provision customer share storage for you.
Cluster resize — change node count (1–4) from the Admin UI Cluster page or CloudFormation; scaling to zero uses stop/start flows (see Stopping the cluster above), not the slider.

Active Directory and identity

Active Directory is required — all SMB access uses domain credentials (Kerberos/NTLM).
Guest and anonymous SMB are not supported.
Single domain per deployment — cross-domain trust is not supported.
Leave Domain is not supported — to join a different domain, use Reset Appliance on the Setup page, then complete First-Time Setup with the recovery secret from Secrets Manager. Customer shares, EFS data, and ACL metadata in DynamoDB are retained.

Workloads and SMB features not supported

Roadrunner SMB is Windows file sharing on EFS — not a full Windows file server. The following are not supported at GA:

Database storage over SMB (for example SQL Server data or log volumes)
VSS / shadow copies and VSS-dependent backup workflows
DFS namespaces and DFS Replication as a native Windows target
SMB Multichannel, SMB Direct (RDMA), SMB compression, and protocol-layer SMB encryption
Named pipes, RPC endpoints, quotas, and persistent handles
Ultra-low-latency local block or HPC scratch workloads

Workloads requiring VSS, DFS, Multichannel, or other deep Windows-native SMB features may be better served by Amazon FSx for Windows® File Server or self-managed Windows file servers.

For the full protocol matrix and EFS/NFS semantic differences, see Known Limitations — SMB Protocol Features.

Security, access, and operations

Admin UI — HTTPS on port 443 via the stack NLB. Restrict with AdminIngressCidr on the nested stack (updates NLB security group rrsmb-*-nlb, inbound 443). Use a VPN or jump-box CIDR for tighter production access — not AdminCidr (that is node metrics/SSH only). See Security overview.
SMB access — clients use NlbDnsName on TCP/445 from networks allowed by SmbClientCidr. Custom DNS names require a CNAME or alias to the NLB.
Monitoring — the Admin UI shows cluster-wide aggregate metrics. There is no built-in CloudWatch, Grafana, or Datadog integration; per-node Prometheus is available on port 9090 for advanced diagnostics.
Support — generate and attach a Support Report from the Admin UI when contacting support (reports are not uploaded automatically).