Levai, Daniel wrote:
> Hi everyone!
>
> I noticed facette became broken in 6.7. Do you guys use any neat tool to
> graph collectd RRDs? Preferably in ports or at most something
> git-pullable but with no outside deps (relative to base or ports).
>
> Thanks for the input!
>
> Dani
Hi Dani,
Before I replay to your question just as a side note. This is not really
OpenBSD related question so please feel free to send me a private email
if you want to discuss this further instead of bothering everyone.
The lack of decent front-end is really achilles tendon of Collectd. Even
the website
https://collectd.org/wiki/index.php/List_of_front-ends
is grossly outdated. The net/librenms port
https://openports.pl/path/net/librenms
is your best bet. LibreNMS is SNMP-based network monitoring system
written in PHP and requires MySQL database. It is a fork of Observium.
https://www.observium.org/
While Observium/LibreNMS primary focus is on monitoring SNMP capable
devices both can can integrate with Collectd to show graphs drawn from
Collectd modules in the web interface.
https://docs.observium.org/collectd/
I use Observium running of Ubuntu 16.04 to monitor a smallish
infrastructure consisting of about 100 devices. I monitor through SNMP
pretty much everything you can think of: batteries on UPSs, electric
consumption of GPU computing nodes via switched PDUs monitoring, managed
switches, servers, named it. In addition to SNMP polling, I also run
Collectd on all my production servers (about 70 Open/Free BSD
infrastructure and Red Hat computing nodes) and push RRDs to the
centralized Observium server. Observium automatically build incredibly
pretty and informative graphs from RRDs. You don't need to build your
own dashboards like with Grafana (IIRC facette requires to build your
own dashboards and it is not even very good). On Red Hat where Collectd
IPMI plugin actually works you also get those nice IPMI graphs.
I tried to run LibreNMS of OpenBSD kernel on bare metal. It was just too
sluggish and web interface was not really usable comparing to Observium
on Ubuntu. I never really bothered to troubleshoot the problem. It could
be that LibreNMS at that time was just not polished enough (they were
much more agressive in adding features than Observium free community
edition). It could be the file system performance. I tried using both
SSD and spinning drives but no big difference. I tried memcached and few
other things. If you search through misc archive with keyword LibreNMS
you will see bunch of my posts and developers/users who were trying to
help.
Observium is officially(per alpha male Adam Armstrong) is designed to
run only designed to run on Ubuntu or Debian and my experience confirms
that. My Observiu/Ubuntu runs as DomU on Alpine Linux Xen Dom0 instance.
I do use raw SSD block devices as a storage. Network interface is 1
Gigabit and even though I have 10 Gigabit card and 10 Gigabit network I
didn't bother to upgrade.
Before I wrap up this long email I would like to bring to your attention
another option. Instead of directly drawing Collectd RRDs you can use
carbon or gmond plugin
https://collectd.org/wiki/index.php/Plugin:Carbon
https://collectd.org/wiki/index.php/Plugin:gmond
to feed collectd into Graphite or Ganglia. I vaguely remember playing
with Graphite plugin but I lost enthusiasm after realizing that I will
have to build by custom dashboards. Jason Dixon who wrote a book on
Monitoring with Graphite should be luring around here so he might be
able to pitch a bit. I never tried running Ganglia on OpenBSD as it is
not in official ports tree.
Just as a final note, if you are doing this to monitor few devices in
your home lab you are way above your head. While all I said is trivial I
would have never done it if somebody was not paying for it. There are
far simpler ways to accomplish above on a small home network.
Best,
Predrag
