[188330] in North American Network Operators' Group
Re: collectd as alternative to RTG for high-resolution polling and
daemon@ATHENA.MIT.EDU (John Kinsella)
Wed Mar 16 15:29:58 2016
X-Original-To: nanog@nanog.org
From: John Kinsella <jlk@thrashyour.com>
In-Reply-To: <CAB69EHjNfaePN0fW+9nC5zJo3rZXLN+ZfWDpKisK_4rZfA9j=g@mail.gmail.com>
Date: Wed, 16 Mar 2016 12:29:53 -0700
To: Eric Kuhnke <eric.kuhnke@gmail.com>
Cc: nanog@nanog.org
Errors-To: nanog-bounces@nanog.org
Collectd is great, IMHO. I was using collectd+graphite to gather and =
display stats for a large collection of VMs, servers, routers, and =
switches. Collectd itself was pretty low overhead, easy to configure (I =
managed configs via puppet) and Just Worked.
Graphite and carbon cache were a little more tricky to set up - carbon =
by default aggregates/averages older data, so if not setup correctly, =
when you go back a few months and try to drill into a graph at a 5 =
minute interval, you get unexpected results.
I=E2=80=99d highly recommend looking at Graphite, as well. Once you get =
used to it, being able to apply functions[1] to aggregate, manipulate, =
and quickly find patterns in data is super useful (ex: look at all =
interfaces on this switch, only display graphs for the top 5 abnormal =
traffic). Jason Dixon has written some great blogs posts about it=E2=80=99=
s use on obfuscurity.com.
John
1: https://graphite.readthedocs.org/en/latest/functions.html
> On Mar 16, 2016, at 11:45 AM, Eric Kuhnke <eric.kuhnke@gmail.com> =
wrote:
>=20
> Would anyone care to share their experience using collectd as an
> alternative to rtg for high-resolution polling of interface traffic =
and
> long term storage?
>=20
> I am investigating the various options for large data set size, =
lossless
> long term traffic charting (not RRAs which lose precision over time). =
One
> possible use is precision 95th billing.
>=20
> https://collectd.org/
