Skip to end of metadata
Go to start of metadata

Purpose

This section describes how Wily can be used to analyze SAP ME system. 

Overview

General

The Investigator component of the Introscope Workstation is the primary interface for troubleshooting production issues. Primary focal points for SAP ME are:

  • Time range – period of time under observation
  • Backends --SAP ME and NetWeaver database metrics and statistics
  • EJB – SAP ME EJB metrics and statistics
  • Frontends = SAP ME application metrics and statistics
  • GC Heap – SAP JVM heap use and garbage collection (GC) information
  • Servlets – SAP ME servlets metrics and statistics
          Select the image to enlarge it.

Time Range

  • Introscope Workstation Investigator allows you to specify time ranges for analyzing the system. ‘Live’ data is displayed by default including 8 minutes prior.
  • You can change the time range to go as far back as there is still Wily data available.
  • Large ‘Time range’ and ‘Resolution’ settings will cause slow response times in the tool. Only include what you need to investigate.
  • Always double-check time range settings as you navigate to different Investigator pages. Values can sometimes get reset back to the defaults.
  • Use ‘Custom’ time ranges to isolate past incidents. For example, if you know an incident occurred in an hour time-frame last week, you should specify that as a ‘Custom’ range rather than pulling back all data collected for the past 7 days.
  • Accept default ‘Resolution’ values as you change ‘Time range’ settings – unless you know you need the granularity.
  • The slider bar is particularly useful for moving time ranges forward or backward.
          Select the image to enlarge it.

Investigator Tab Views

  • The views that appear in the right pane of the Investigator vary, depending on the resource or metric selected in the Investigator tree.
  • Depending on the type of node selected, different tabs are available containing context specific information about the node.
  • For SAP ME troubleshooting, these tab views tend to be the most useful:
    • Overview
    • General View
    • Traces View
  • The Traces View is particularly useful and when available for a selected node, lists the recorded Transaction Trace events for the node. SAP ME requests (usually JSP calls) are displayed and can be sorted by Duration to help determine if any transactions are taking particularly long to complete.
     
         Select the image to enlarge it.
  • The Traces Summary View pane displays a sortable list of calls made for a selected transaction to further isolate bottlenecks.
          Select the image to enlarge it.
  • The TracesTree View is displays the call stack for a selected transaction where duration is represented graphically.
  • This makes it easy to spot calls that are taking too long to complete.
  • You can also select individual calls in the stack and further detail is displayed about the call, including actual SQL statements from recorded backends calls.

    Investigator Tree

  • Backends
    • Backends in the Investigator Tree contains separate nodes for NetWeaver and SAP ME databases.
    • The Overview tab shows performance metrics for each so you can quickly assess whether there are performance issues in either database.
       
           Select the image to enlarge it.
    • You can select the SQL node under the SAP ME database node for additional information about performance of specific SAP ME database calls.
    • The lists can be sorted and you can further drill down for performance metrics for individual SQL statements.
       
           Select the image to enlarge it.
  • EJB
    • EJB node in the Investigator Tree contains information about SAP ME EJBs.
    • Most EJBs in SAP ME are Session beans.
    • You can select individual EJB nodes in the Investigator Tree and the Overview tab displays performance metrics for the individual EJB.
       
           Select the image to enlarge it.
    • You can further drill down for metrics about individual EJB methods.
  • Frontends
    • Frontends in the Investigator Tree contains separate nodes for applications deployed in NetWeaver.
    • When you select an individual application, the Overview tab displays performance metrics for the application. SAP ME Web Application is the main application deployed for the SAP ME solution.
    • Frontends node is good for obtaining performance metrics for an entire application where you can also get more detailed information from the URLs node.
    • For SAP ME, however, the Servlets node contains similar detail information and is often easier to navigate than URLs node.

     Select the image to enlarge it.

GC Heap

  • The Overview tab for the GC Heap node in the Investigator Tree graphically shows heap use.
  • In production, SAP ME heap utilization graph displays the typical sawtooth pattern associated with garbage collection - where Minor GCs occur over time, but overall heap utilization keeps growing over time until the garbage collector initiates a Full GC.
  • This is the normal pattern, but it is often confused as indicating a memory leak.
  • It is only a problem if heap utilization keeps growing over time after Full GCs.
  • The GC Heap graph is very helpful for identifying memory spikes.
  • However, because you cannot differentiate Minor GCs from Full GCs by looking at the graph, log files with verbose GC information are often more useful if you suspect a memory leak.

     Select the image to enlarge it.

Servlets

  • Servlets in the Investigator Tree contains separate nodes for servlet in the container.
  • The Overview tab shows performance metrics for individually selected servlets.
          Select the image to enlarge it.
  •  You can drill down to obtain additional performance metrics for individually selected Servlet nodes.
  • Responses Per Interval is often useful when troubleshooting SAP ME to determine if there was sudden spike in activity for a specific servlet.
          Select the image to enlarge it.

Related Documents

Performance Metrics Monitoring with Introscope by Wily

  • No labels