Hi again,

Not really shure here..

daemonize: false
aggregate: src_as, dst_as
plugin_buffer_size: 2048
nfacctd_port: zzzz
nfacctd_ip: xx.yy.zz.xx
! nfacctd_time_secs: true
! nfacctd_time_new: true
plugins: memory 

I don’t use any sql right now, 

And I get the negativer results when doing this.

pmacct -c dst_as -N xxxx -r ; sleep 300 ;pmacct -c dst_as -N xxxx -r ; sleep
300; pmacct -c dst_as -N xxxx -r
-1458316866
-170266441
-445279125

/Tobias

-----Ursprungligt meddelande-----
Från: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] För Paolo Lucente
Skickat: den 21 mars 2005 11:37
Till: [email protected]
Ämne: Re: [pmacct-discussion] Counter size

Hey Tobias,

On Mon, Mar 21, 2005 at 09:59:53AM +0100, Tobias Bengtsson wrote:

> I wonder how big the byte counter can be, because I have problems with 
> negative counter values..

On pmacct side (that is, we are talking about the cache in the SQL plugin)
counters are 32 bit (at this propo, anyone has either an opinion or comments
about the topic 'u_int64_t & portability' ?); they need to accomodate
counters just for a single 'sql_refresh_time' timeframe. They work the
following way: when the counter is about to hit the upper 32 bit burden,
it's 'archived' and a new counter is started. The next purging event (that
is, when the cache is purged) the actual counter and the archived ones are
framed in distinct SQL queries (and rewritten as long unsigned integers -
%lu).

Said this all, the negative counters should be generated by the DB. It could
be matter of just enlarge the field to some more bytes. With such volumes,
did you enable 'sql_history' ? By framing counters in fixed timeslots (say,
5-10 minutes or even 1 hour), you will be able to push a reliable upper
burden to the growth of the counters into the DB.

Cheers,
Paolo


_______________________________________________
pmacct-discussion mailing list
[email protected]
http://muffin.area.ba.cnr.it/mailman/listinfo/pmacct-discussion


Reply via email to