That sounds like it could be a signed/unsigned counter wrapping issue.
This happens especially on counter metrics that store large values; network
IO metrics are common offenders. There are a few "remove peaks/spikes"
scripts floating around that may be of some help.
On Mon, Nov 23, 2015 at 4:41
Hi all,
I am seeing a strange behaviour on the aggregated summary plots whereby I see
negative spikes that do not appear on any of the underlying clusters. These
spikes can be quite frequent (every other pull...)
Has anyone seen this before?
Thanks
Cumprimentos / Best regards,
Cristóvão José Do
2 matches
Mail list logo