Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

Planning

This section covers the following topics.

Prerequisites

You must meet the following prerequisites before commencing setup.

Deployed cluster infrastructure

Ensure that your Amazon networking requirements and Amazon EC2 instances where SAP workloads are installed, are correctly configured for SAP. For more information, see SAP NetWeaver Environment Setup for Linux on Amazon.

See the following ASCS cluster specific requirements.

Supported operating system

Protecting ASCS with a pacemaker cluster requires packages from SUSE, including targeted cluster resource agents for SAP and Amazon that may not be available in standard repositories.

For deploying SAP applications on SUSE, SAP and SUSE recommend using SUSE Linux Enterprise Server for SAP applications (SLES for SAP). SLES for SAP provides additional benefits, including Extended Service Pack Overlap Support (ESPOS), configuration and tuning packages for SAP applications, and High Availability Extensions (HAE). For more details, see SUSE website at SUSE Linux Enterprise Server for SAP Applications.

SLES for SAP is available at Amazon Web Services Marketplace with an hourly or annual subscription. You can also use the bring your own subscription (BYOS) model.

Required access for setup

The following access is required for setting up the cluster.

Reliability

The SAP Lens of the Well-Architected framework, in particular the Reliability pillar, can be used to understand the reliability requirements for your SAP workload.

ASCS is a single point of failure in a highly available SAP architecture. The impact of an outage of this component must be evaluated against factors, such as, recovery point objective (RPO), recovery time objective (RTO), cost and operation complexity. For more information, see Reliability in SAP Lens - Amazon Well-Architected Framework.

SAP and SUSE references

In addition to this guide, see the following references for more details.

You must have SAP portal access for reading all SAP Notes.

Concepts

This section covers Amazon, SAP, and SUSE concepts.

SAP – ABAP SAP Central Services (ASCS)

The ABAP SAP Central Services (ASCS) is an SAP instance consisting of the following two services. It is considered a single point of failure (SPOF) in a resilient SAP system.

SAP – Enqueue Replication Server (ERS)

The Enqueue Replication Server (ERS) is an SAP instance containing a replica of the lock table (replication table).

In a resilient setup, if the standalone enqueue server (EN/ENQ) fails, it can be restarted either by restart parameters or by high availability software, such as Pacemaker. The enqueue server retrieves the replication table remotely or by failing over to the host where the ERS is running.

Amazon – Availability Zones

Availability Zone is one or more discreet data centers with redundant power, networking, and connectivity in an Amazon Web Services Region. For more information, see Regions and Availability Zones.

For mission critical deployments of SAP on Amazon where the goal is to minimise the recovery time objective (RTO), we suggest distributing single points of failure across Availability Zones. Compared with single instance or single Availability Zone deployments, this increases resilience and isolation against a broad range of failure scenarios and issues, including natural disasters.

Each Availability Zone is physically separated by a meaningful distance (many kilometers) from another Availability Zone. All Availability Zones in an Amazon Web Services Region re interconnected with high-bandwidth, low-latency network, over fully redundant, dedicated metro fiber. This enables synchronous replication. All traffic between Availability Zones is encrypted.

Amazon – Overlay IP

Overlay IP enables a connection to the application, regardless of which Availability Zone (and subnet) contains the active primary node.

When deploying an Amazon EC2 instance in Amazon, IP addresses are allocated from the CIDR range of the allocated subnet. The subnet cannot span across multiple Availability Zones, and therefore the subnet IP addresses may be unavailable after faults, including network connectivity or hardware issues which require a failover to the replication target in a different Availability Zone.

To address this, we suggest that you configure an overlay IP, and use this in the connection parameters for the application. This IP address is a non-overlapping RFC1918 private IP address from outside of VPC CIDR block and is configured as an entry in the route table or tables. The route directs the connection to the active node and is updated during a failover by the cluster software.

You can select any one of the following RFC1918 private IP addresses for your overlay IP address.

If, for example, you use the 10/8 prefix in your SAP VPC, selecting a 172 or a 192 IP address may help to differentiate the overlay IP. Consider the use of an IP Address Management (IPAM) tool such as Amazon VPC IP Address Manager to plan, track, and monitor IP addresses for your Amazon workloads. For more information, see What is IPAM?

The overlay IP agent in the cluster can also be configured to update multiple route tables which contain the Overlay IP entry if your subnet association or connectivity requires it.

Access to overlay IP

The overlay IP is outside of the range of the VPC, and therefore cannot be reached from locations that are not associated with the route table, including on-premises and other VPCs.

Use Amazon Transit Gateway as a central hub to facilitate the network connection to an overlay IP address from multiple locations, including Amazon VPCs, other Amazon Web Services Regions, and on-premises using Amazon Direct Connect or Amazon Client VPN.

If you do not have Amazon Transit Gateway set up as a network transit hub or if it is not available in your preferred Amazon Web Services Region, you can use a Network Load Balancer to enable network access to an overlay IP.

For more information, see SAP on Amazon High Availability with Overlay IP Address Routing.

Amazon – Shared VPC

An enterprise landing zone setup or security requirements may require the use of a separate cluster account to restrict the route table access required for the Overlay IP to an isolated account. For more information, see Share your VPC with other accounts .

Evaluate the operational impact against your security posture before setting up shared VPC. To set up, see Shared VPC – optional.

Pacemaker - STONITH fencing agent

In a two-node cluster setup for a primary resource and its replication pair, it is important that there is only one node in the primary role with the ability to modify your data. In the event of a failure scenario where a node is unresponsive or incommunicable, ensuring data consistency requires that the faulty node is isolated by powering it down before the cluster commences other actions, such as promoting a new primary. This arbitration is the role of the fencing agent.

Since a two-node cluster introduces the possibility of a fence race in which a dual shoot out can occur with communication failures resulting in both nodes simultaneously claiming, “I can’t see you, so I am going to power you off”. The fencing agent is designed to minimise this risk by providing an external witness.

SLES supports several fencing agents, including the one recommended for use with Amazon EC2 Instances (external/ec2). This resource uses API commands to check its own instance status - “Is my instance state anything other than running?” before proceeding to power off its pair. If it is already in a stopping or stopped state it will admit defeat and leave the surviving node untouched.

Pacemaker - simple-mount and classic architecture

This guide covers two architectures for SAP cluster solutions on SLES for SAP – simple-mount and classic (previous standard). Simple-mount was certified as the SLES for SAP Applications cluster solution in late 2021. It is now the recommended architecture for both ENSA1 and ENSA2 deployments running on SLES for SAP 15 and above. For more details, see SUSE blog Simple Mount Structure for SAP Application Platform.

If you are configuring a new SAP installation, we recommend the simple-mount architecture. If you already have the classic architecture, and wish to migrate to the simple-mount architecture, see Switching architecture to simple-mount.

The following are the differences between the classic and simple-mount architectures.

See the Architecture diagrams for more details.

Automation and Amazon Launch Wizard for SAP

You can set up a cluster manually using the instructions provided here. You can also automate parts of this process to ensure consistency and repeatability.

Use Amazon Launch Wizard for SAP for automated deployments of SAP NetWeaver, SAP S/4 HANA, SAP B/4HANA, and Solution Manager. Launch Wizard uses Amazon CloudFormation scripts to quickly provision the resources needed to deploy SAP NetWeaver and S/4 HANA. The automation performs SAP enqueue replication and pacemaker setup so that only validation and testing are required. For more information, see Amazon Launch Wizard for SAP.

To ensure that the behavior and operation of your cluster is well understood regardless of how your system is set up, we recommend a thorough test cycle. See Testing for more details.

Switching architecture to simple-mount

Follow along these steps if you want to switch an existing cluster with classic architecture to use the recommended configuration of simple-mount architecture.

These steps must be performed in an outage window, allowing stop/start of services and basic testing.