Hi Linus,

Inline:

On Thu, Jan 17, 2013 at 04:46:58PM +0100, Linus Nordberg wrote:

> Is there an easy way to dump the internal RIB of nfacctd? I'm trying to
> understand why AS_PATH keeps being '^$' for all the entries we see.

No - although the feature is on the roadmap. For debugging purposes, so
temporarily until all is solved, in the configuration you can: 1) set
bgp_daemon_msglog to true, 2) set debug to true aswell and 3) send all
stdout output to a log file (logfile directive).

> [ .. ]
> 
> For info, SRC_AS and DST_AS are both 0 for all entries when
> nfacctd_as_new is set to 'bgp', which I guess indicates that nfacctd
> realises that its BGP info is bad or nonexistent. (Setting
> nfacctd_as_new to 'fallback' gives good result, presumably based on
> netflow data.)

Correct. Essentially i grasp all BGP-related primitives (src_as, dst_as
and as_path) on which you are aggregating are being left empty, "^$" being
an internal pmacct representation for empty, when sourcing data from BGP.
Is the IP address being used to export NetFlow the same of either the BGP
router-ID or the BGP source-address? If not, that explains: the correlation
is not made possible automagically - you might opt for a bgp_agent_map map;
if all is good that side of the things then you should inspect what is
actually sent to pmacct via BGP - with the bgp_daemon_msglog described
above. Give it a try and let's take it from there.

> BTW, we're using mongodb, which seems to work just fine. Happy to follow
> up with more details on how this scales once we're up and running in
> production. (That said, we've tested all of this with the memory plugin,
> to ensure that this is not a backend issue.)

Great stuff. Sure, more than interested in receiving scalability figures
and/or comparisons with other backends.

Cheers,
Paolo


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

Reply via email to