I have noticed that LibreNMS is polling incorrect CPU Processor Load Percentage for Juniper EX4600 Switch. It will fluctuate for normal to an extremely high CPU Percentage upward in 80 to 90%.
I have a Juniper JTAC Case opened and they are comparing LibreNMS Graphs to output of their Juniper CPU/CPU Threads script that is being run every minute. And JTAC is saying the highest spike we had was 68% and no 90% as Libre shows. They also said that snmp process on the Juniper was using 2% CPU and mib2d was at 8% CPU, so snmp polling contributed 10% to the output according to the CPU script being run on the Juniper.
Hmm I’ve seen similar with an EX series where I get a high cpu alert, but when I check it on the device there was no load showing in the process list at all. Load wise the port polling tends to be quite slow on the junipers for some reason.
I think he want you to run the poller manually, in debug, and check the Processor module output, to see which OIDs are being used by LibreNMS when polling. Then one can confirm whether the OIDs are indeed the correct ones. ./poller.php -h DEVICE_ID -d
@murrant LibreNMS is definitely not showing the correct CPU Usage.
I connected a test EX4600 switch to my lab and added it to LibreNMS. LibreNMS shows that CPU Usage nearly got to 60% usage and right now it shows to be at 29%.
However, in command line on the Juniper, I am running the following command to get real time CUP Usage and piping to refresh it every 2 seconds - “show snmp mib get 1.3.6.1.4.1.2636.3.1.13.1.8.9.1.0.0 | refresh 2”
And command line on Juniper shows the max it gets to in real time is 17%.