Different CPU values reported in SNMP polling from the actual values on VCO monitor

Products

• VMware VeloCloud SD-WAN

Issue/Introduction

Customer may notice high CPU values or sometimes equals to 0 when polling edges and VCGs OIDs:

cpu_300s_avg_pct              .1.3.6.1.4.1.45346.1.1.2.2.2.4
cpu_60s_avg_pct                .1.3.6.1.4.1.45346.1.1.2.2.2.5

- "debug.py --health_report" output also shows higher CPU percentage or 0.

root@IN-54-5S-gateway-1:~# debug.py --health_report | grep -i cpu
  "cpu_300s_avg_pct": "33.3",
root@IN-54-5S-gateway-1:~# debug.py --health_report | grep -i cpu
  "cpu_300s_avg_pct": "33.3",
root@IN-54-5S-gateway-1:~# debug.py --health_report | grep -i cpu
  "cpu_300s_avg_pct": "33.3",
root@IN-54-5S-gateway-1:~# debug.py --health_report | grep -i cpu
  "cpu_300s_avg_pct": "50.0",
root@IN-54-5S-gateway-1:~# debug.py --health_report | grep -i cpu
  "cpu_300s_avg_pct": "0.0",
root@IN-54-5S-gateway-1:~# debug.py --health_report | grep -i cpu
  "cpu_300s_avg_pct": "0.0",
root@IN-54-5S-gateway-1:~# debug.py --health_report | grep -i cpu
  "cpu_300s_avg_pct": "33.3",
root@IN-54-5S-gateway-1:~#  

- VCO monitor page would reflect stable CPU values:

Environment

VCE and VCG builds 5.2.3.x

Cause

SNMP polling depends on debug.py --health_report command which is never intended to be executed in loop.

Issue is calling load average in short intervals is the problem, and this is similar to defect#148624 which is expected to be fixed in future builds: Yamazaki 6.0.1.0, Zarza 6.1.0.0 and Woodford 5.2.6.0. In fixed images customer will need to use "cpu_usage_percent" in polling rather than "cpu_300s_avg_pct".

 

Resolution

Fix will be available in future builds starting from Yamazaki 6.0.1.0, Zarza 6.1.0.0 and Woodford 5.2.6.0.

Workaround: is to increase the interval in which this command is getting called like running script every 30 seconds or 1 minute. And if possible, you may do it every few minutes rather than seconds.