Hello,

Yes, it is trickier with the aggregator.. the pcap syntax is really 
powerful and definetely is a must to have an option to read the pcap and 
aggregate filters from a file. I am not sure about CPU load though.. If 
the aggregator consumes less CPU than pcap filters, it is good to have a 
reverse option... I haven't tested which one (aggregate_filters or 
networks_file aggregator) is less stressful to the CPU under heavy loads 
(over 100Mbits/s or 15000 packets/s).

Paolo Lucente wrote:
> i see your point. And was wondering whether a better solution could be
> to add a feature which allows the filter directives (aggregate_filter,
> pcap_filter) to read the well-formatted pcap filter from a file - like:
> 
> aggregate_filter: 'file:/path/to/a/filter'
> 
> It could come handy to generate and update filters (which could be
> read at the next startup of the daemon).
> 
> The point here is that the solution should be generic in order to be
> useful: a flat file that lists some prefixes just eases your life to
> either include or exclude them from somewhere - that's it; a filter
> expression is more featureful (which is not necessarily a good thing,
> as it also translates in added complexity) but fits in pretty much
> every scenarios.
> 
> The networks_file directive is an aggregator rather than a pure filter
> (it does a bit of filtering though). Excluding prefixes then poses the
> problem of what to do with the IP addresses that would pass the filter
> - should they get aggregated somehow?
>>
>> It would be nice to have an option to exclude lots of networks from
>> accounting -> just the opposite of networks_file. It is not very practical
>> to write/generate huge pcap filters if you want to exclude 100 or more
>> networks from pmacct.

-- 
Anton Glinkov
network administrator

_______________________________________________
pmacct-discussion mailing list
http://www.pmacct.net/#mailinglists

Reply via email to