Skip to Content.

rare-users - Re: [gn4-3-wp6-t1-wb-RARE] rare testing....

Subject: RARE user and assistance email list

List archive


Re: [gn4-3-wp6-t1-wb-RARE] rare testing....


Chronological Thread 
  • From: Simon Leinen <>
  • To: mc36 <>
  • Cc: <>
  • Subject: Re: [gn4-3-wp6-t1-wb-RARE] rare testing....
  • Date: Sun, 7 Mar 2021 17:34:51 +0100
  • Authentication-results: mx1.switch.ch; x-trusted-ip=pass

mc36 writes:
> and even more fun in the attached picture, the iperf 1day test ended
> imho in 2 hours ago so grafana started showing us 70tbps peaks,
> telling me that tofino can exceed it's datasheet 6tbps limit... :))))

Presumably an arithmetic error involving counter roll-overs or some
other unexpected effect of limited precision. Might be worth a look.
Such measurement artifacts can be very annoying, e.g. if they cause
false alarms in a monitoring system.

Is this using SNMP counters natively? (Because those have a pretty
well-defined semantic.)

Or does Grafana get its data from Prometheus? Prometheus decided to only
use IEEE 754 double-precision floats, which allow for 53 bits of integer
precision (which presumably gets worse and worse over time as the
exponent grows...).
--
Simon.



Archive powered by MHonArc 2.6.19.

Top of Page