[mrtg] Fwd: Re: Minimum hardware

[Freddie Cash]

---------- Forwarded message ----------
From: "Freddie Cash" <fjwcash at gmail.com>
Date: Jun 1, 2016 9:06 PM
Subject: Re: [mrtg] Minimum hardware
To: "thomaz portella" <thomazp at gmail.com>
Cc:

I run a monitoring server using 2x Xeon CPUs (single core) and 7 GB of RAM,
with 4x 76 GB SCSI disks in hardware RAID10, running Debian 7, MRTG,
Routers2, Nagios, and LibreNMS.

MRTG polls around 300 hosts, graphing all NICs, disks, CPUs, RAM, load
averages, and a few other one-offs like SMTP queues. Everything stored in
rrd files via rrdcached. Some things like NICs are polled every minute.
Some things like CPU and RAM are polled every 5 minutes. Others like
storage usage every 15 minutes. Depends on how often the thing being polled
changes.

Routers2 does the displaying.

Nagios does all the host, network, connectivity, and service monitoring for
the above hosts, polling every 5 minutes, including another 100 or so
devices.

And we've started to move things into LibreNMS for more in-depth
monitoring, polling just about everything with a MIB every 5 minutes
(currently 524 hosts/devices).

We're just starting to hit disk and CPU bottlenecks on that box. I'm in the
process of moving it to an 8-core Opteron box with 32GB of RAM and 8 SATA
disk in software raid10. Once that is done, MRTG/Routers2 will be replaced
completely by LibreNMS and another 1000+ switches will be added.

If that supports the load, then we'll look into adding approx 5000 Linux
stations, spread across 40-odd locations, as well. At that point, we may
consider using the server in each building as a distributed poller, getting
everything back via rrdcached and memcached.

It all depends on how many targets you'll be querying and how fast your
disk setup is that determines how much hardware is needed. :)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.oetiker.ch/pipermail/mrtg/attachments/20160601/2f391200/attachment.html>