Hello Paolo,

Thank You for answer - I'm using version 1.5.0 but I checked it with
version from cvs too and the problem was exactly the same.
I've done different test - replicate traffic received from junipers to
loopback and from there I used flow-tools to send
traffic to nfacctd - the results were the same. It seems that nfacctd where
traffic is received is making some problems.
In another step I split configuration:
in my original nfacctd file I left all other data sources - this is my
config:

daemonize: true
debug: false
pidfile: /var/run/nfacctd.pid
syslog: daemon
!

aggregate: tag, src_as, dst_as, peer_src_as, peer_dst_as, peer_src_ip,
peer_dst_ip, in_iface, out_iface

nfprobe_version: 10
nfacctd_ip: x.x.x.x
nfacctd_port: 2077
nfacctd_time_new: true

nfacctd_as_new: fallback

nfacctd_net: fallback
nfacctd_disable_checks: true
nfacctd_renormalize: true

pre_tag_map: /etc/pmacct/pretag.map
sampling_map: /etc/pmacct/sampling.map

plugins: mysql[DATA1], mysql[DATA2], mysql[DATA3], mysql[DATA4],
mysql[DATA5]
plugin_pipe_size: 4096000
plugin_buffer_size: 4096
pre_tag_filter[DATA1]: 701
pre_tag_filter[DATA2]: 501
pre_tag_filter[DATA3]: 301
pre_tag_filter[DATA4]: 711
pre_tag_filter[DATA5]: 901

bgp_daemon: true
bgp_daemon_ip: x.x.x.x
bgp_daemon_max_peers: 10
bgp_peer_src_as_type: bgp
bgp_src_as_path_type: bgp

sql_optimize_clauses: true
sql_dont_try_update: true
sql_multi_values: 1024000
sql_db: pmacctdb
sql_host: localhost
sql_passwd: very_secure_password
sql_table_version: 7
sql_table_type: bgp
sql_cache_entries: 256000

sql_history_roundoff[DATA1]: m
sql_history[DATA1]: 1m
sql_refresh_time[DATA1]: 60
sql_table[DATA1]: acct_bgp_DATA1_IN_%Y%m%d_%H
sql_table_schema[DATA1]: /etc/pmacct/schema/acct_bgp_data1.schema

for DATAx sql settings are exactly the same only table and schema file is
different - and one of the flows which is replicated I configured in second
config file:

daemonize: true
debug: true
pidfile: /var/run/nfacctd_2.pid
syslog: daemon
!

aggregate: tag, src_as, dst_as, peer_src_as, peer_dst_as, peer_src_ip,
peer_dst_ip, in_iface, out_iface

nfprobe_version: 5
nfacctd_ip: y.y.y.y
nfacctd_port: 2088
nfacctd_time_new: true

nfacctd_as_new: fallback

nfacctd_net: fallback
nfacctd_disable_checks: true
nfacctd_renormalize: true

pre_tag_map: /etc/pmacct/pretag.map
sampling_map: /etc/pmacct/sampling.map

plugins: mysql[DATA10]
plugin_pipe_size: 4096000
plugin_buffer_size: 4096
pre_tag_filter[DATA10]: 801

bgp_daemon: true
bgp_daemon_ip: y.y.y.y
bgp_daemon_max_peers: 10
bgp_peer_src_as_type: bgp
bgp_src_as_path_type: bgp

sql_optimize_clauses: true
sql_dont_try_update: true
sql_multi_values: 1024000
sql_db: pmacctdb
sql_host: localhost
sql_passwd: very_strong_password
sql_table_version: 7
sql_table_type: bgp
sql_cache_entries: 256000

sql_history_roundoff[DATA10]: m
sql_history[DATA10]: 1m
sql_refresh_time[DATA10]: 60
sql_table[DATA10]: acct_bgp_DATA10_IN_%Y%m%d_%H
sql_table_schema[DATA10]: /etc/pmacct/schema/acct_bgp_data10.schema

it's almost exactly the same - the difference is only in nfprobe_version
and IP addresses of course - hard to say but maybe there is some problem
with nfprobe version
but in first config I have configured nfprobe_version: 10 and receive flows
from version 5 and data is counted correctly

After I split configuration in two files everything is working fine

Before split I even checed data in sql - made simple sql query to count
bytes from table and there was a difference - when two flows was send to
nfacctd sum of bytes was
lower then there was send only one flow

In this case I can say there is no problem with flows replication -
currently tee is working fine - nfacctd is suspect ;) but logs are clear,
no error, no packet drops on interface,
small load on machine - any ideas how to find the issue ?

Best regards
Adam Bogdan


2015-05-13 18:26 GMT+02:00 Paolo Lucente <[email protected]>:

> Hi Adam,
>
> Reading about your problem, i'm at a loss too. Can you please further
> refine the issue by looking, say with tcpdump, what gets replicated?
> Maybe we discover one of the two plugins stops teeing alltogether? What
> pmacct version are you using?
>
> Cheers,
> Paolo
>
> On Thu, May 07, 2015 at 04:49:48PM +0200, Adam Bogdan wrote:
> > Hi,
> >
> > I have some odd problem
> > This is what I have - Juniper router with logical-systems - I have some
> > routes (BGP sessions) in primary routing table (non LS) and different
> > routing table in one logical-system.
> > I set up a BGP sessions from pmacct to both "routers" (non-LS and LS).
> > Juniper export all flows with IP address from non-LS - because I need to
> > resolve flows based on BGP in LS I'm doing this like this:
> > flows are send to IP e.g. a.a.a.a on port 3000 from there I replicate
> flows
> > to 127.0.0.1 to ports 5000 and 6000
> > Then I run two tee plugins with this configuration:
> > nfacctd_port: 7000
> > nfacctd_ip: 127.0.0.1
> >
> > plugins: tee[lo5]
> >
> > tee_receiver[lo5]: b.b.b.b:2001
> > tee_source_ip[lo5]: c.c.c.c
> > tee_transparent[lo5]: false
> >
> > and second config:
> > nfacctd_port: 6000
> > nfacctd_ip: 127.0.0.1
> >
> > plugins: tee[lo6]
> >
> > tee_receiver[lo6]: b.b.b.b:2001
> > tee_transparent[lo6]: true
> >
> > Small explanation for above - flows from Juniper are replicated to ports
> > above (5000 and 6000) and from them I send them to pmacct (b.b.b.b) - for
> > lo5 I change IP address for IP from LS (c.c.c.c) and for lo6 I leave it
> > unchanged (IP from non-LS)
> > now on pmacct machine I get two exactly the same flows but visible as
> sent
> > from two machines (to this point everything looks fine, I even checked
> > packets send from Juniper to tee and then sent to pmacct (iptables
> > counters) and it looks fine)
> >
> > The problem - when I enable only tee[lo5] I get proper traffic value on
> > pmacct but when I enable tee[lo6] then traffic which I get in graphite
> > instantly drop
> > Here You can see how it looks like: http://postimg.org/image/zb9u1ywaj/
> > To 21:00 I get some traffic (enabled tee[lo5] and tee[lo6]) after 21:00 I
> > disabled tee[lo6] and traffic instantly increased to the proper value -
> > after 22:00 enabled tee[lo6] again
> >
> > The problem is exactly the same if I enable/disable tee[lo5] - then on
> > tee[lo6] traffic value increase or deacrease
> >
> > I'm sitting on this second days and no idea where to search - nfacctd
> > config file on pmacct machine should be ok - beacause it's working for
> > others flows which I get from other routers
> >
> > This is screen from today: http://postimg.org/image/6qgd4sztj/ - after
> > 16:00 I enabled one of tee
> > In pmacct logs there are no errors - I even enabled debug for one of
> > plugins:
> > May 07 16:43:01 DEBUG ( DATA1/mysql ): 975 VALUES statements sent to the
> > MySQL server.
> > May 07 16:43:01 INFO ( DATA1/mysql ): *** Purging cache - END (PID: 2502,
> > QN: 975/975, ET: 0) ***
> > Only when I get data from sql - there is a big difference between 15:54
> and
> > 16:08
> >
> > Anyone have any idea where to search ?
> >
> > Best regards
> > Adam Bogdan
>
> > _______________________________________________
> > pmacct-discussion mailing list
> > http://www.pmacct.net/#mailinglists
>
>
> _______________________________________________
> pmacct-discussion mailing list
> http://www.pmacct.net/#mailinglists
>
_______________________________________________
pmacct-discussion mailing list
http://www.pmacct.net/#mailinglists

Reply via email to