IT-Conductor Blog

SAP HANA Distributed System - Guide to Scale Out, HA and Monitor HANA

Posted by Terry Kempis on Oct 31, 2016 7:54:42 PM

 Setting up Scale Out, High Availability and Monitor HANA

HANA scale out is unlike other database platforms, it can be simple but complex at the same time.  Depending on the scenario, the in-memory nature of HANA services, the persistence layers and optional multi-tenancy (
+

SAP Patches Can Fix or Break SAP Performance

Posted by Linh Nguyen on Oct 5, 2016 10:55:45 PM


INTRODUCTION:

Ever hopeful or worried that SAP patches may fix or break your system's performance?  Let's face it, if we have a specific problem, we hope a patch will fix it.  However, given patches are now mostly delivered in packages or bundles (particularly support packs), we're worried at the same time it may break something else.  Even if all functionality remains intact, there is always a chance performance may be impacted, for better or worse.

That's why we recommend to include SAP Performance Best Practices for Implementation, Upgrade & Migration.  Case in point, while monitoring Solution Manager with IT-Conductor (yes that will be another topic in itself that we'll cover in another post), we were pleasantly surprised, unlike most cases, that a set of patches actually improved the system performance.  Without the proper SAP monitoring tools and context, it would be pretty hard to pinpoint the cause and effect of such action.

Let's see how we became aware of problems, then related the cause and effect:

1. The Problem and Symptom

  • Service Level Alerts and notification indicated that the system was under stress with automated Health checks
  • We also received email notification via subscription that one of the app servers had prolonged periods of 'CPU Idle 0%'

2. Find the Root-cause Part 1

  • In the Health chart with RED status bars above, we can click on a single button to determine the cause from all underlying components monitored for the system (or what we better refer to as a Service)
  • Expanding the service tree where there are warning indicators we find both Background utilization and CPU Idle on the same page with a high degree of negative correlation (i.e. Batch processes maxing out while CPU Idle hugs the zero line)

3. Find the Root-cause Part 2

  • What workload was running that took all the batch work processes.  Well this is pretty hard to find in hindsight, but we've got a secret weapon!  Click on the app server icon having the issues and a menu pops up with a choice to look at WorkProcesses snapshot
  • Looks like we had a combination of Solman batch and ChaRM running a long time, and no free batch processes

4. The Fix

  • The Basis team applied some Solman add-on support packages, and guess what happened - Look at the charts below - the middle chart shows when patches were applied, things began to improve around mid-night.  Was it coincidence?  Maybe, but what's probable, look at the correlation when charts of metrics and events are synchronized with each other, all on a single-pane

+
FREE IT-Conductor Trial to Monitor SAP

Posts by Topic

see all

Recent Posts

Subscribe to Email Updates