Siteminder Policy Trace Analysis


SMPolicyTraceAnalysis is a Support Tool for analysis of various Siteminder trace logs and reporting and graphing of the results.


The program is the evolution of various prior perl and python scripts into a java program which then generates graphs and PDF reports of the results.  Most recently, it has been expaned to include the "Stats" analysis, and can now also parse Web Agent Trace logs.

When run it looks like this:


Screen Shot




Overview


SMPolicyTraceAnalyis has three major report types.   

  1. Policy Trace Log for smtracedefault.log analysis. 
  2. Stats Analysis for extracting ad graphing the "Statistics" lines out of smps.log logs.
  3. Agent Trace Log for analysis of agent trace logs.
Each of the sections are configurable and is explained in more detail below.

There is always more that can be done, if you have problems or suggestions then please do email me, Mark O'Donohue <odoma04@ca.com> .


Other places to check for updated information are :

Marks blog page (occasional explaination of updates)

Good Luck - Mark

Policy TraceLog Analysis

Graphs and reports begin/end transactions, and various other statistics, from the detailed trace logs.

Some examples of the major reports are :

  1. Bar Chart of Transaction response times (as well as table)




  2. Graph showing number of concurrent threads processing transations.



  3. Chart showing total time spent by all threads between each trace print line (generally maps to time spent in the function)





  4. Transactiobns over cuttof time now highlight slow steps in colour gradient.






  5. Ability to edit/add the pattern matching and also edit the graph report labels and titles.




  6. As well as the PDF report, the images are generated seperatly, in case you want to include them in a seperate report/email.





Policy Stats Analysis

Before you begin with Policy Server Stats Analysis, you need to setup of the batch script or cron job, click this line for more details. 


Policy Server statistics reports rely on the regular generation of the "Statistics" line in the smps.log files.  From those it is possible to graph the connections, threads and throughput of messages over a day, week or month and it gives a good insight into the health of the policy server.

The "Stats" analysis can be generated regularly via script (DOS or sh) and best practice is to setup a cron/schedule job to have them always generated on a regular basis.

Once a Cron or batch job is setup regularly issuing "-stats" commands you can then generate the reports.


Some of the reports are :


  1. Policy Server Connections over time




  2. Policy server Message throughput over time



  3. Internal Message Queue Depth over time

    Although this graph does not look so interesting, the queue depth > 1 is always an indication of some delay occuring, and it often corresponds to when problems occur in processing transactions.





  4. Number of active threads over time




WebAgent Trace Log Analysis

The graphs for the agent trace logs are much the same as the policy server transaction logs, just derived from the agent log with mappings for start and end transations.  Two examples are:

  1. Transaction Time




  2. Number of transactions started and completed per sec.




Contact: