On Sat, Jul 20, 2019, 9:24 AM Joe Clarke (jclarke) <jcla...@cisco.com> wrote:
> > > > On Jul 20, 2019, at 06:46, Eliot Lear <l...@cisco.com> wrote: > > > > > > >> > >> I see you’re using a 32-bit int for the drop-count. Would it not make > sense to make this a 64-bit counter instead? Yeah, this number should be > low, but if something goes crazy, having a larger field space might be > useful. > > > > It’s a lot of drops. Let’s talk about scaling this because there are a > great many. > > It is, yes. But if you started to see a lot of this, you might be able to > identify a compromise. > A relative drop count is probably sufficient to identify issues. For example, if the manufacturer is informed that packets to local network are being dropped (his device wants to send to a locally bound controller) then it probably means that the controller has not been bound. Exact drop counts are probably not required. I think it would be worthwhile to run through a complete scenario (i.e. simulated "malfunctioning" device and diagnostic report) to determine what is actually required to be reported. Ranga > Joe > > _______________________________________________ > OPSAWG mailing list > OPSAWG@ietf.org > https://www.ietf.org/mailman/listinfo/opsawg >
_______________________________________________ OPSAWG mailing list OPSAWG@ietf.org https://www.ietf.org/mailman/listinfo/opsawg
