Hi Anatoly,

On Tue, Aug 23, 2011 at 11:14:16AM +0400, Anatoly Bashmakov wrote:

> As far as I understand, stamp_inserted and stamp_updated are fields
> related to when data go to db. stamp_inserted fields are the same in one
> time-bin. For what I'm asking are for example flowStartSeconds[1] and
> flowEndSeconds[2].

It depends: if nfacctd_time_new is set to true then stamp_inserted does
refer to when data is received at the collector; otherwise the original
timestamp received from the probe, ie. flowStartSeconds, is used. Whereas
stamp_updated refers to the last time a certain record is touched; the
likes of flowEndSeconds are mostly ignored unless nfacctd_sql_log is set
to true, ie. creation of time-bins is disabled, causing timestamps up to
the second resolution to be used to populate stamp_inserted/stamp_updated
fields. Lastly, hoping the probe doesn't provide more than a single time
source for a certain template, nfacctd is able to use flowStartSeconds,
flowStartMilliseconds or flowStartSysUpTime as indication - if multiple
are defined, one is picked in a deterministic way, ie. flowStartSysUpTime
wins over flowStartMilliseconds which in turns wins over flowStartSeconds.
Hope this helps.

> For probe we're using Juniper MX240 with JunOS 10.4R5.5 and MPC-2-3D.
> pmacct from cvs, fetched month ago works flawlessly!

Thanks for confirming that.

Cheers,
Paolo


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

Reply via email to