[smokeping-users] Rasp-Pi smokeping slaves
gregs at sloop.net
Wed May 1 01:44:26 CEST 2013
>> Perhaps slightly off-topic, but I thought it might be interesting to
>> others and a starter for ideas for others...
>> Building on the concept of using a Raspberry Pi as a smokeping slave
>> [Thanks to Ioannis Theodoridis] for the push!] I've completed one.
JZ> This is an AWESOME idea and I would love a copy of the code.
JZ> We just bought 2 Pis for this purpose.
JZ> If we decide to run one as a master do they still handle the load and the config ok?
JZ> Thanks a ton!
I tested a few different configs: [These were all the the "B" version
Pi with 512M RAM.]
1) Smokeping/MRTG/Nagios - as a stand-alone master. [With Smokeping
hitting say - 30-40 devices/60 secs with fping, MRTG polling a hundred
SNMP devices/ports/60secs and Nagios handling essentially the same
hosts as Smokeping - mostly for notification purposes. [I actually had
nagios watching the SP RRD files and not doing it's own checks.]
In this case, the Pi was pretty buried. If you OverClocked it to 900
Mhz [IIRC] it would just barely stay below a 1.0 load average - at
least some of the time. But it was very touch and go - I never had a
problem, but when the 15 min load avg is really close to 1 and
sometimes over, then it's likely you're going to have problems at some
point. Since this is a critical warning/monitoring system, I'm not so
hip on running them on the ragged edge. :)
2) Running it as just a smokeping and MRTG master - same number of
targets etc. This works great and load is much less. I'm not sure why
Nagios is such a pig, but it is...
3) Running smokeping as a slave over an SSH tunnel back to the
mothership. And running MRTG as a "master" [not that there's any other
way to run MRTG] on the Pi, storing all the MRTG data locally on the
SD card, and generating all the graphs locally etc.
Again, this is absolutely no problem with any semi-reasonable number
of targets. [Say less than 100 smokeping and 100 MRTG targets.]
And with smokeping "phoning" home to give the data to a master, I can
run nagios there and generate all the alerts from a central location.
Load averages, even when cranking MRTG graphs [which only occur when
you're viewing pages with routers2.cgi] is really no problem at all.
Load is nearly always very low, or well below 1.
So, I tossed my configs and have it setup as an image, with option 3.
I have a few tweaks I need to make to that image, but other than that
it's ready for anyone to use.
It would be quite trivial to take that image and make smokeping a
master on if that's what you want - as everything it needs is already
I'm not sure exactly how/where to place the image so more than a
single person or two could pull it - it's about 2G.
Before running it, you'll want to write it on at least a 4G card [if
you're going to store much on the SD card, say running MRTG with big
RRD's, like I do] and you'll want to resize the partition. [Easy with
I run in on the RPi-Debian distro.
Finally, I've tried to secure it pretty substantially - blocking all
inbound ports except http/81 [can't use 80, since smokeping uses it to talk to the
master] https/443 and ssh/22.
I've setup apache to do auth for connects to MRTG too.
So, with the image - it's literally about 15 minutes to setup a full,
secure SSH tunnel smokeping slave from start-to-finish.
MRTG setup with say a router or two, and a switch is another few
minutes - depending on how much massaging you need to do the MRTG
config files. [I have some great sed scripts to hack those up quick in
most of the cases I have to use it in...though those aren't part of
Suggestions, questions, thoughts? I'm glad to answer.
More information about the smokeping-users