[collectd] Re: New plugin "email" to collectd ham, spam, ... statistics

[Staněk Luboš]

Hi Sebastian,
now we can see the threading solution in real. :)

Sebastian Harl napsal(a):
> This plugin collects email count and size for each type (e.g. ham, spam,
> virus, ...) of emails, spam score values and the count of successful spam
> checks (e.g.  BAYES_99, SUBJECT_DRUG_GAP_C, ...).
> 
> These information are provided by external programs which communicate with the
> plugin thru a UNIX socket and a simple line-based protocol:
> 

Could you provide some example to start from?
I am currently trying to accommodate 'swatch' but I am not fluent in
perl scripting and log parsing.

> /* e-mail type (e.g. ham, spam, virus, ...) and size */
> e:<type>:<bytes>
> 
> /* spam score */
> s:<value>
> 
> /* successful spam checks */
> c:<type1>[,<type2>,...]
> 

Could we split it to incoming/outgoing?

I am also using a 'greylist' filter.
Could we also collect greylist's connection checks
(SPF-compliant/autowhitelisted/delayed) or something similar?


> At most MAX_CONNS (currently set to 5) clients can connect to the plugin
> simultaneously. Each connection is handled by a separate thread. Any input
> line is limited to 256 characters (including the newline character) which
> ought to be enough for anybody[tm] by definition. "c"-lines have to be split
> up if they grow longer.
> 
> Signed-off-by: Sebastian Harl <sh<span style="display: none;">.trailing-username</span>(a)<span style="display: none;">leading-domain.</span>tokkee.org>

I think that the line buffer limit is not big enough.
I am using my own special SpamAssassin rules with longer names and also
several SARE rules (http://www.rulesemporium.com/) and the list of spam
checks can be really long. I checked my maillog for several days and the
list of checks approximates to 230 with spikes over 290.
No surprise, I am not anybody[tm]. :)

Best regards,
Lubos