Hi Sophie,

Let me start with the bad news to conclude with the good ones.

Unfortunately there is not a good way to size memory pools given a
traffic figure and/or the amount of IP addresses monitored. It really
depends on the traffic mix (that is, how big it is the <src_host,
src_port> matrix produced by your traffic?). So unless you are in a
constrained / embedded environment my recommendation would be to set
imt_mem_pools_number to zero so to let the memory table grow without
boundaries; then, once you have some estimate, if you wish you may
define some boundaries to go in a more production phase.

Alternatively you could resort to a push model, like the print plugin
(and Kafka, RabbitMQ and SQL ones). In those data is pushed at regular
time intervals and shall cache get full (symptom you may want to config
a bigger cache), entries are dumped before the next scheduled write. 

On Mon, Dec 24, 2018 at 01:50:32PM +0100, Sophie Loewenthal wrote:
> Hi,
> I just installed pmacct and would like to assign the memory 16Mb.
> I read these pages,
> https://github.com/pmacct/pmacct/blob/master/docs/INTERNALS
> and know that this should be adjusted : 
> imt_mem_pools_number 
> imt_mem_pools_size
> I’m storing,
> aggregate: etype, proto, src_host, src_port
> The config file has this,
> daemonize: true
> pidfile: /var/run/pmacctd.pid
> syslog: daemon
> aggregate: etype, proto, src_host, src_port
> interface: ens3
> daemonize: true
> aggregate: sum_host
> What should the values be for imt_mem_pools_size? 
> How many pools should I have?
> The server has incoming 6 Mbytes/s a second during peak periods from some 
> 3000 IP addresses.
> Regards,
> Sophie 
> _______________________________________________
> pmacct-discussion mailing list
> http://www.pmacct.net/#mailinglists

pmacct-discussion mailing list

Reply via email to