Hi Rashid, Thanks for raising your issues.
I’ve tried to tackle each of your points in order. 1) Immediately after setup, the Bono Client count is 30 (on average) even before I registered any numbers. When the SIPp stress test node is started, this client count reduces to 0. If my understanding of client count as "active calls" is right, then this doesn't seem right. That is strange. I think the stat that you are referring to is the number of TCP connections which clients have tried to register over, so I would expect it to start at 0. We should be able to work out what is going on with some more diagnostics. Firstly, can you send me the output of running snmpwalk to query that statistic, and the command that you ran, to make sure that we are talking about the same one. Second, can you send me the output of netstat -lp when the extra clients have shown up? That should allow us to check whether the TCP connections really exist. Thirdly, can you please turn on debug logging<http://clearwater.readthedocs.org/en/latest/Troubleshooting_and_Recovery.html?highlight=debug%20log#bono> on your Bono node, and send the Bono logs from /var/log/bono? This will help us to debug if there is an issue with the code. 2) What is the meaning of bono latency, sprout latency, and sprout PRR ? PRR also reduces when the system is stressed! The permitted request rate (PRR) is an estimate of how much load the system can handle. As Sprout takes on more load, it is able to build a better estimate of how much more it can handle, and this will cause it to lower the PRR. This is detailed internal information, so it’s likely that you will not need it. You can find details of how our load monitoring and overload control work in two blog posts: · http://www.projectclearwater.org/clearwater-performance-and-our-load-monitor/ · http://www.projectclearwater.org/overload-control-2/ Bono and Sprout both provide statistics covering latency of SIP requests handled by this node over the given period in an SNMP table. The table contains statistics for average latency, variance, highest latency, and lowest latency. Latency is measured in microseconds. Our article on statistics<http://clearwater.readthedocs.org/en/latest/Clearwater_SNMP_Statistics.html#usage> explains how to read these tables. 3) All the Chronos related statistics do not show up. First, we should make sure that you have all of the necessary packages installed to get these statistics. You can ensure this by running sudo apt-get install clearwater-snmpd on each node, and sudo apt-get install clearwater-snmp-handler-astaire on your Sprout and Ralf nodes. Second, how are you querying these statistics? Are you using the snmpget and snmptable commands as described in http://clearwater.readthedocs.org/en/latest/Clearwater_SNMP_Statistics.html? If so, can you send the exact commands you ran, and what the output was? 4) Is there a statistics which captures failed registrations/calls? You can find out the number of failed registrations in each time period by looking in the sproutInitialRegistrationTable. It contains counts of the number of attempts, successes, and failures of initial registrations. There are similar tables for re-registrations and de-registrations. You can find out the number of failed calls in each time period by looking in the sproutSCSCFIncomingSIPTransactionsTable. It contains the number of attempts, successes, and failures of incoming SIP transactions for the SCSCF. The number of failed INVITE transactions will be the number of failed calls. I hope that helps. Yours, Chris From: Clearwater [mailto:[email protected]] On Behalf Of Rashid Mijumbi Sent: 14 March 2016 06:18 To: [email protected] Subject: [Project Clearwater] Meaning of Cacti Graphs Dear all, I am seeking guidance on the meaning of project cleawater's SNMP statistics. I have a manual installation with 6 VM nodes one for each of sprout, bono, ellis, homer, homestead and ralf. I could not find an explanation for the Cacti graphs which show the following: 1) Immediately after setup, the Bono Client count is 30 (on average) even before I registered any numbers. When the SIPp stress test node is started, this client count reduces to 0. If my understanding of client count as "active calls" is right, then this doesn't seem right. 2) What is the meaning of bono latency, sprout latency, and sprout PRR ? PRR also reduces when the system is stressed! 3) All the Chronos related statistics do not show up. 4) Is there a statistic which captures failed registrations/calls ? Thank you, Rashid
_______________________________________________ Clearwater mailing list [email protected] http://lists.projectclearwater.org/mailman/listinfo/clearwater_lists.projectclearwater.org
