collectd - Unable to configure "sensors"

Setting up an Enterpise type monitoring system, I am unable so far to monitor metrics from openSUSE using collectd.
Am able to setup on CentOS compiling collectd from source. Compiling from source can be important, it is the only way to make plugins available in collectd. When collectd is compiled, it looks for libaries it recognizes that relate to plugins and then automatically compiles support for the plugin as well.

Unfortunately, compiling collectd from source on openSUSE does not seem to find the libraries so at the moment I am attempting to use the pre-built collectd package from the OSS repo.

  1. Trying to use the sensors plugin.**
    Inspecting the collectd configuration file (/etc/collectd.conf, I find that sensors plugin support has been compiled but disabled. Enabling the plugin results in the error “file not found” and indeed, there is no sensors code, library and config in the following default (and configured) plugins location
ls /opt/collectd/lib64/collectd/

As I described, AFAIK creating these missing files requires collectd source.

NOTE: This suggests that the collectd config file has been simply copied from the Builder’s machine without the plugins, and whether the plugin code themselves might be copied could be a complex question in itself.

2. Trying to use the ipmi plugin
Not knowing much about ipmi, I tried it anyway. Don’t know if it supports hardware.


But, this results in the following error

Jan 28 12:19:51 ES-installed collectd[13556]: ipmi plugin: ipmi_smi_setup_con failed: No such file or directory
Jan 28 12:19:51 ES-installed collectd[13556]: ipmi plugin: thread_init failed.

And searching the system for ipmi_smi_setup_con turns up empty (maybe it’s a method or procedure and not a file?)

So, specific Questions:

  1. If anyone has gotten collectd to run collecting metrics from either of the two plugins I tried, was wondering how you did it.
  2. If anyone has an idea how to collect system metrics in another way, would like to hear about any alternatives. Collecting metrics is only part of the issue, I’d also like a method for sending the metrics to an aggregator… amost any method would be acceptable by either pipelining or TCP/IP (eg netcat could be a useful part of the solution).


For anyone who is looking to collect raw system metrics in an Enterprise environment, currently collectd is mostly broken.
I have created a bug

Hopefully it’ll get fixed possibly implementing one of my suggestions in the bug.

In the meantime as of today, I’m starting to look at some kind of solution using a local system analysis tool like siga in combination with sending it across the network using something like netcat.

Life is more difficult when ready-made solutions aren’t working…