Hi Sven, What are the specs of the box? Running high on CPU? How much memory? And what would approx be the average traffic mix size? It looks quite strange that only 750mbps of traffic let such big buffers to run out of space. Usually a ~10MB pipe size with 10KB buffer size, configured as follows, is enough:
plugin_pipe_size: 10240000 plugin_buffer_size: 10240 Please let me know. Cheers, Paolo On Tue, Mar 03, 2009 at 01:08:59PM +0100, Sven Schmitt wrote: > Hi, > > i am running pmacct 0.11.4 on FreeBSD 7.1 for traffic accounting in a > network with about 10.000 client ip addresses and a bandwidth of about > 750 mbit/s. > > I got a lot of these messages: > > Mar 2 15:00:00 host1 pmacctd[37092]: - increase shared memory size, > 'plugin_pipe_size'; now: '161920000'. > Mar 2 15:00:00 host1 pmacctd[37092]: - increase buffer size, > 'plugin_buffer_size'; now: '161920'. > Mar 2 15:00:00 host1 pmacctd[37092]: - increase system maximum socket size. > Mar 2 15:00:00 host1 pmacctd[37091]: ERROR ( in/memory ): We are missing > data. > > > My pmacct configuration: > > debug: false > daemonize: true > syslog: kern > interface: em4 > aggregate[in]: dst_host > aggregate[out]: src_host > networks_file: /usr/local/etc/networks.def > plugins: memory[in], memory[out] > imt_path[in]: /tmp/in.pipe > imt_path[out]: /tmp/out.pipe > imt_mem_pools_number: 0 > imt_mem_pools_size: 16384 > plugin_pipe_size: 161920000 > plugin_buffer_size: 161920 > > A cronjob runs "pmacct -s -p /tmp/out.pipe -e" and "pmacct -s -p > /tmp/in.pipe -e" every 5 minutes to fetch the collected data. > > When i increase the values for plugin_pipe_size and plugin_buffer_size > pmacct becomes instable and dies from time to time. > > Do you have any suggestions what to optimize? > > Best Regards > Sven Schmitt _______________________________________________ pmacct-discussion mailing list http://www.pmacct.net/#mailinglists
