Page Comparison

...

 Below is a diagram of how the load test works.  The Wanpipe driver is configured for API mode, where an application called BERT is used to run the load test through the attached Sangoma hardware.

Load test scenario

Below are the steps required in order to run a load test

...

1. In a separate terminal screen run the ifconfig command and monitor for ‘overruns’.  If the ‘overruns’ counters for your Wanpipe ports increment  during the load test, this means that your server is not able to efficiently handle the data sent from the ports to your kernel.

   The following command will provide you a live output of ‘ifconfig’ so that you can monitor overruns in real time:

   -> watch –n 1 “ifconfig |grep “overruns””
   

   The above output shows low amount of overruns.  This is perfectly fine during the first few seconds as everything is trying to sync up.  If the counters continue to increment every second, there is a problem.

   If you experience overruns while running the Wanpipe load test, your entire issue is system related (and not the lines plugged into the system). Run the steps indicated on this page: http://wiki.sangoma.com/troubleshooter-boards-pri-audio-issues PRI- Audio/Noise Issues

2. Run the Linux vmstat command to monitor the system CPU usage during the performance test.  Run the following line to watch the output in real time

   -> vmstat –n 1
   

   

   The numbers in the red outline above indicate the entire CPU (all cores) idle, per second.  You do not want your server idle CPU usage to go below 60% during the production test.  Keep in mind that you will possibly be adding other applications to this system, so anything lower than 60% system idle may cause load issues in the future.  However, if you are not adding any application, and your system item goes down to 20%, your system may perform fine. Simply monitor the rest of the items listed here to conclude system sanity.
    

3. Monitor the /var/log/messages file during production test.  If there are any issues relating to hardware, overruns, NMI, and physical layer, they will display in this file.  If there are any messages that present themselves that appear somewhat concerning, please contact Sangoma support at support.sangoma.com or visit our online documentation for troubleshooting at Sangoma Knowledge Base

...