Weird spikes on interfaces

[jodok.glabasna (Jodok)]

An update on my investigations about some users seeing spikes:

1. one user found out that his spikes correlate with the checkmk server running on 100% CPU usage at that time the spike occurred.
   He assigned more CPUs to the machine and is now fine.

2. Two users provided debug log data that shows that the devices (cisco and juniper) really reported (wrong) high in and out octet counters.

Looks like there are buggy switch/router devices or at least buggy firmwares out there that report wrong in/out octets from time to time:

• https://www.reddit.com/r/Juniper/comments/gg6fg6/juniper_qfx5100_snmp_polling_issue_highincorrect/

• https://www.zabbix.com/forum/zabbix-troubleshooting-and-problems/439582-zabbix-5-4-8-false-high-bandwidth-monitoring

3. We have a lot of Partners that have customers with thousands of Ports/Interfaces that never see/saw such spikes.

I do not think any more that there is a problem with checkmk.
Perhaps if you upgrade your switch/router to latest firmware you are fine.