[c-nsp] VSS and 3560, convergence measurement
Geert Nijs
geert.nijs at gmail.com
Tue Nov 16 16:50:30 EST 2010
When i was measuring VSS convergence, i actually noted that it is better to
use LACP or Pagp control protocol in active mode instead of putting a
portchannel in fixed status "on".
My explanation for this was: when a portchannel uses LACP or Pagp active
mode, it only starts forwarding from the moment BOTH sides have agreed on
the portchannel settings.
I have noticed that when both sides are ON, one side immediately starts
sending packets from the moment the interface comes up. The other side might
not yet be ready to forward packets,
therefore it drops packets. This brought down packet drops from 1.5 seconds
to 100 msec for me :-)
regards,
Geert
2010/11/15 Pavel Skovajsa <pavel.skovajsa at gmail.com>
> Hello Mkhail,
>
> Usually VSS convergence should be below 200ms, at least on paper.... I
> would try to turn off various "automatic features", starting with
> trunk negotiation (DTP) on one end of spectrum and undesirable
> portchannel on the other to find the culprit.
>
> -pavel
>
> On Sun, Nov 14, 2010 at 8:25 PM, Mikhail <hidden.node at gmail.com> wrote:
> > Hello, list.
> >
> > We have VSS 1440 and we try to measure convergence time when we pull one
> of
> > the copper links from 3560 (Etherchannel, PAgP, desirable-desirable, we
> also
> > tried on-on).
> >
> > We do it by sending "ping -f -i 0.1 -W 0.1 <IP>" from the Linux station:
> >
> > Linux -- [3560] === [VSS] === [4948] --- <IP>
> >
> > This command prints us about 12-18 dots which is 1200-1800ms (btw, when
> we
> > pull one of the 10Gb links from 4948 it prints usually 1-2 dots,
> sometimes 3
> > dots).
> >
> > VSS design guide states that for 3xxx series port failure convergence
> time
> > should be about 200ms.
> >
> > 1. Does "ping" measurement actually give us convergence time? Is there
> any
> > "official" method for such type of measurement?
> >
> > 2. I suspect that 3560 is the root of the issue. Is it normal for that
> > switch to converge for 1,5 second after losing one of the Etherchannel
> > links?
> >
> > Thank you in any advance!
> >
> > _______________________________________________
> > cisco-nsp mailing list cisco-nsp at puck.nether.net
> > https://puck.nether.net/mailman/listinfo/cisco-nsp
> > archive at http://puck.nether.net/pipermail/cisco-nsp/
> >
>
> _______________________________________________
> cisco-nsp mailing list cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/
>
More information about the cisco-nsp
mailing list