Hi Stefano, I recommend to change the relationship between the two buffer sizes: try to get from 1:2000 to, firstly, 1:1000 then to 1:100 and see if that improves the situation.
Cheers, Paolo On Tue, Nov 27, 2012 at 07:03:17PM +0100, stefano birmani wrote: > Hi, > > we've a pmacct installation on a RHEL Server rel 5.5, (Linux XXX > 2.6.18-194.26.1.el5 #1 SMP XXX x86_64 x86_64 x86_64 GNU/Linux) and > following is our nfacctd.conf file: > > daemonize: true > logfile: /usr/local/pmacct/var/nfacctd.log > nfacctd_disable_checks: true > nfacctd_time_new: true > nfacctd_renormalize: true > nfacctd_net: netflow > nfacctd_port: 5678 > nfacctd_ip: XXX.XXX.XXX.XXX > plugins: mysql[pni] > plugin_buffer_size: 819200 > plugin_pipe_size: 1638400000 > sql_optimize_clauses: true > sql_dont_try_update: true > sql_cache_entries: 9999991 > sql_multi_values: 900000 > sql_history_roundoff[pni]: m > sql_history[pni]: 5m > sql_refresh_time[pni]: 300 > sql_table[pni]: acct_pni_%Y%m%d_%H%M > sql_host[pni]: 10.0.0.2 > sql_table_schema[pni]: /usr/local/pmacct/etc/acct_pni.sql > bgp_daemon: true > bgp_daemon_ip: XXX > bgp_daemon_max_peers: 500 > bgp_table_peer_buckets: 100 > bgp_neighbors_file: /usr/local/pmacct/var/routers.txt > bgp_aspath_radius: 3 > bgp_agent_map: /usr/local/pmacct/etc/agent.map > nfacctd_as_new: bgp > aggregate[pni]: in_iface, peer_dst_as, peer_src_as, peer_src_ip, peer_dst_ip > bgp_peer_src_as_type: bgp > pre_tag_map_entries: 1000 > sampling_map: /usr/local/pmacct/etc/sampling.map > refresh_maps: true > > Whenever we restart the daemon after a while we start seeing in the > log file multiple entries like the following: > > XXX ERROR ( pni/mysql ): We are missing data. > XXX If you see this message once in a while, discard it. Otherwise > some solutions follow: > XXX - increase shared memory size, 'plugin_pipe_size'; now: '1638400000'. > XXX - increase buffer size, 'plugin_buffer_size'; now: '819200'. > XXX - increase system maximum socket size. > > We tried, as suggested, to change both the plugin_pipe_size and the > plugin_buffer_size values increasing them until the values shown but > nothing has changed... > > We've found that when the message is printed in the log file the > difference between the seq values is always equal to the ratio between > the plugin_pipe_size and the plugin_buffer_size values (2000 in our > case), for example: > > ((struct ch_buf_hdr *)rg->ptr)->seq: 5808 seq: 7808 > > Can anyone help us? We need pmacct :-) so before starting studying the > src code we ask for help because we don't know how to go on. > > Thanks, > Stefano > > _______________________________________________ > pmacct-discussion mailing list > http://www.pmacct.net/#mailinglists _______________________________________________ pmacct-discussion mailing list http://www.pmacct.net/#mailinglists
