Re: [munin-users] Best Practices suggestions for good plugin graphs

[Jerome W. <jwa...@be...>]

On Wed, 29 Feb 2012 15:41:04 +0100, Jerome Warnier wrote:
> I really wonder if this is the right place to discuss such a topic.
> Please tell me in case it's not.

I'm already answering to myself over some of these topics.

> So, I carefully read http://munin-monitoring.org/wiki/plugin-bcp.
>
> I personally feel that there are some important subjects not covered:
> 1) The documentation chapter should be more detailed, IMHO, there
> should really be the following items in the documentation 
> (preferrably
> in the file itself): Author (and an e-mail address),
> version(s)/changelog, supported platforms (or at least the ones known 
> to
> work).

> 2) It should be suggested somewhere that Perl is the preferred 
> language
> (for integration, maintenance, portability, debugging and performance
> reasons), immediately followed by POSIX shell. Anyone willing to 
> write a
> new plugin from scratch would better consider using Perl. Then shell
> scripts (trying to avoid bash specifics) and finally anything else
> (Python, Ruby, ...).

I added a link from http://munin-monitoring.org/wiki/HowToWritePlugins 
to http://munin-monitoring.org/wiki/ShellPlugins in the meantime. Covers 
parts of it.

> 3) It is indeed suggested here
> (http://munin-monitoring.org/wiki/HowToWritePlugins) to avoid forking
> subprocesses whenever possible (again, for portability, debugging and
> performance reasons).

Although, http://munin-monitoring.org/wiki/ShellPlugins should also 
suggest it.

> 4) There should be a way to measure plugin performance on a given
> system. Maybe a new tool basically doing a munin-run while "timing()"
> it? Or maybe something in the plugins themselves...

> 5) There are many plugins implementing memory usage tracking for
> processes. Though they seem each to do it their own way (which is not
> portable most of the time, some using various flavours of ps, others
> using pmap, others yet using top, ...). I think these should be
> rewritten to all use the same way, preferably implemented in the 
> common
> plugin infrastructure.

A good starting point might prove to be this effort: 
http://munin-monitoring.org/wiki/PluginFramework.

> Let me know your thoughts.
Hope it helps.