marcus_qc said:
Constant low traffic vs bursty high traffic.
This. My OVHcloud graph looked like a seismograph until I moved backups to 3am. Now it's flat except for the nightly spike which I expect.
436 days. reboot is surrender.
My bandwidth graph looks like a heartbeat, what's wrong?
uma said:
Moved backups to 3am
3am in which timezone? My Contabo in Singapore hits peak local evening when EU wakes up. Timezone math is the real ops work.
one small ping for man...
everseoul
Member OP
armstrongvds said:
Timezone math
Found it. Contabo monitoring agent reporting to Nuremberg every 2h. Killed it, graph flat. Thanks all, was driving me insane.
DANA3
Member
everseoul said:
Killed it
You killed monitoring to fix a graph? Bold. I would have just filtered the agent IP in Grafana. The state file remembers all.
DANA3 said:
Filtered the agent IP
Or just rate the interface differently. Most panels have a min interval setting. 2h spikes disappear at 4h zoom.
Following this. My Hetzner auction box does the same thing but every 6h. Never found the cause.
marcus_qc
Member
lowwatch said:
Every 6h
Check for apt unattended-upgrades. Default on Debian images, hits mirrors on a timer. 6h is common for the random sleep spread.
42U and still growing
marcus_qc said:
apt unattended-upgrades
Can confirm. My Mexico City box to Hetzner Falkenstein mirror, 6h spikes. Changed to daily in apt config. Problem solved.
marchhopper
Member
20 years and I still forget to check cron first. Every. Single. Time.
marchhopper said:
Forget to check cron first
Because it's always cron. Or DNS. Or MTU. The holy trinity.