Auto Scaling Solution

This repository provides an approach and sample code for auto scaling SAP application servers in Azure based on SAP performance metrics.

• Solution Overview
• Setup Instructions
• Additional UseCases

Disclaimer

This is just a prototype to demonstrate how auto scaling of SAP application servers can be achieved in Azure. The approach and the code needs to be thoroughly tested and modified as per requirements before it can be deployed in a productive SAP environment. Also, operational aspects like error handling, security etc. needs to be considered before using it in a production SAP environment.

Solution Overview

Autoscaling solution discussed here is split into 2 parts

SAP Telemetry Collection

SAP AutoScaling Solution

SAP telemetry collection

SAP telemetry collection architecture is as shown below. In this architecture Logic App is used to ingest the SAP performance metrics data from /SDF/MON_HEADER (or /SDF/SMON_HEADER depending on what is scheduled) into a custom log table in log analytics workspace.

The Logic app is scheduled to run peridically with a recurrence trigger to pull the data from SAP based on time filter and push it into Log analytics workspace. For example if /SDF/MON collection is scheduled to run every min and logic app recurrence interval is set to 5 min, each execution of Logic app workflow will pull the data for the past 5 min.

SAP AutoScaling Solution

Below is an overview of the SAP App server Scale in and Scale out architecture.

SAP Application Server Scale Out Architecture

• Using Azure Monitor create a log alert which queries the Custom log table with the SAP performance data and creates an alert when specific threshold is breached.
• The alert triggers a Azure Automation Powershell runbook. The runbook checks the Scaling Config table to see if current app server count is equal to max. app server count. If yes it exits without performing any action.
• If not it triggers an ARM template to create a new application server based on Custom VM Image (of a pre-built app server).
• After successful creation of app server, the runbook calls Logic app to regiser new application server in SAP Logon and Server groups. Groups information are fetched from Config table.
• Finally the PS runbook updates the Current app server count in the Scaling Config table.

SAP Application Server Scale in Architecture

• Using Azure Monitor create a log alert which queries the Custom log table with the SAP performance data and creates an alert when specific threshold is breached.
• The alert triggers a Azure Automation Powershell runbook. The runbook checks the Scaling Config table to see if current app server count is equal to min. app server. If yes it exits without performing any action.
• If not the runbook calls Logic app to to de-register the application servers from Logon and Server groups. Group information is fetched from the Config table.
• The runbook then schedules a second runbook based on Delay timeout defined in the config table. This allows exising user sessions/jobs to be drained from the application server before its Stopped.
• The second runbook issues a SoftShutdown Command on the application server with a timeout fetched from config table.
• Once the SAP application server is successfully stopped, the corresponding Azure resources (VM, NIC and disks) are deleted from within the runbook.
• Finally the PS runbook updates the Current app server count in the Scaling Config table.

Setup Instructions

• Steps to deploy the SAP telemetry solution discussed above using Terraform can be found here
• Steps to deploy the Autoscaling solution discussed above using Terraform can be found here

Additional UseCases

• Custom Monitoring Solution - Incase you would like to use a different monitoring and alerting solution (like solution manager) you can still use the above framework for auto scaling.

  • Import the automation automation runbooks, ARM templates and shell script located within [here] (autoscaling/scalingartifacts) in the scaling artifacts directory.
  • Create a webhook for the automation runbook and call it from your monitoring/alerting solution based on required threshold.

• Periodoc Scaling - Runbooks can be used to schedule scaling for seasonal elasticity (month-end, peak trading period etc.). Increment size for scaling is configurable (i.e. add 'n' app servers for month-end)