On Wed, 2011-04-13 at 20:08 +0100, Gavin McCullagh wrote:
> On Wed, 13 Apr 2011, Rick Jones wrote:
> 
> > I'd suggest checking the sflow "flow" with tcpdump but there are still
> > some holes there.  At first blush that sounds like something is off by a
> > byte or four somewhere.  Certainly I've had a few of those while I've
> > been trying to flesh-out tcpdump's support of sflow printing.
> 
> I've looked at it, but not that sure what to look for.  Is there a spec
> somewhere I can compare to?

http://www.sflow.org/ is a place to start with sflow.  If you want to
use tcpdump, let me know offline and I'll send you my latest
print-sflow.c

rick

> 
> > > Any thoughts on what to do?
> > 
> > ulimit -c unlimited?-)  At least that is what I've done on the off
> > chance I'll get another crash.  
> 
> Tried that.  No core that I can find.  I stuck it in /etc/default/ntop
> (which is sourced into the init script).  I then ran ntop manually with it
> set and no luck either.
> 
> > Thusfar I'm just getting the "attempt to write a second time with the
> > same timestamp" messages from RRD.  (where I suspect if the interval is
> > ever just a skosh less than one second that will happen... although I
> > thought I saw it at least once when I had the "store every" settings
> > above one second)
> 
> I tried to increase the sample rate to 10000 so I didn't get more than one
> report per second.  Then, on the console I ran it on, ntop stays up for a
> while (probably waiting for the first sample), then I get a gazillion lines
> of output all saying:
> 
> flow_sample_element length error (expected 0, found 4)
> flow_sample_element length error (expected 0, found 4)
> 
> I'll keep looking, thanks.
> 
> Gavin
> 


_______________________________________________
Ntop mailing list
[email protected]
http://listgateway.unipi.it/mailman/listinfo/ntop

Reply via email to