please see comments inline.
regards
.siva
>
>
> well, i've had experience with dot1q on a pa-fe-fx on a vip2-40(64m)
> and dot1q on a geip+. under both 12.0(13-17)S and 12.1(7)E.
>
> for the pa-fe-fx, you can type "ip route-cache distributed" under
> 12.0()S and it sticks, but it doesn't actually make any difference in
> the cpu utilization on the rsp. going to 12.1(7)E fixed that.
sounds like DCEF didn't get enabled for some reason.
would you happen to have opened a TAC case or gathered any information
from this event?
some of the things i would be looking for would include
sh ver
sh run
sh cef line detail
sh cef int
sh diag
sh int
sh int stat
sh controller [fast | gig] X/Y/Z
sh proc cpu
>
> so i swapped out the vip2-40/pa-fe-fx for a geip+. net throughput
> dropped from 90-92 mbps (under normal conditions) to 75 mbps (there's
> a news server on the far end of the switch). sporadic packet loss.
> running a 1pps ping with 1460 byte packets to a host on a different
> vlan on the same interface (not the news server) about 98% of the
> packets made it through ok. when packet loss happened it was usually
> two packets at a crack, and then it was ok for a while. lather,
> rinse, repeat.
makes sense :( if you are losing packets, your TCP sessions go into
slow start, so you have less traffic to send. why you are losing
packets, especially on a periodic basis is kinda strange.
need same info as above
>
> i'm currently back on 12.0(17)S, with the news server non-trunked on
> an extra geip+ (the on-site spare). this cuts back drasticly on the
> amount of traffic on the trunked geip+, so no packet loss (between the
> same two hosts as before).
>
> have dcef turned on on both geip+es, but the cpu utilization is still
> noticeably higher than when i was running dot1q on a pa-fe-tx. there
> are some folks in nz that i'm aware with who have a case open with the
> tac on a very similar problem (talking to a 2948g).
shouldn't be :(
again, if you have a TAC case or the info i requested above, it's
something i can look at.
>
> so now i'm kind of torn; i'm not sure i want to go playing with
> 12.1()T unless there are specific representations that dot1q problems
> have been addressed. word on the street is that 12.2 should be kept
> away from production equipment (which would have been my instinct
> anyway).
i am surprised at this :( 12.2 is only bug fixes from 12.1()T.
there are no new features. admittedly 12.2 is less stable then 12.1
which is less stable then 12.0. but 12.0 has had a longer period of
feature freezing compared to 12.1 which has had 8 maintenance
builds (compared to 1 for 12.2) might be useful to figure out if
these commands are made with respect to 12.0 / 12.1 or with 12.1()T.
afaik 12.2 is more stable then 12.1()T [with some exceptions].
>
> thoughts? you're always the voice of reason and clue here siva, want
> to venture a guess on this one?
can't figure it out. if you can provide me the information or TAC
case numbers i will be happy to scrub it this week. (you can
send me the info offline). i will try my best to scrub the info / cases,
but i also need to clean up networkers presentation, before leaving for
LA the coming weekend. anybody going to be at LA/Chicago? :) come say hi!
cheers
.siva
>
> ---rob
>
>
> Siva Valliappan <svalliap@cisco.com> writes:
>
> > possibly because dot1q is only DCEF switched in 12.0()S / 12.1()T / 12.2 ?
> >
> > cheers
> > .siva
> >
> >
> > On Sat, 16 Jun 2001, Arie Vayner wrote:
> >
> > > Hi
> > >
> > > What is the problem with dot1q and GEIP+?
> > >
> > > Arie
> > >
> > > -----Original Message-----
> > > From: rs@seastrom.com [mailto:rs@seastrom.com]
> > > Sent: Thursday, June 14, 2001 20:34
> > > To: Karyn Ulriksen
> > > Cc: cisco-nsp@puck.nether.net; rs@valhalla.seastrom.com
> > > Subject: Re: [nsp] GEIP vs. GEIP+
> > >
> > >
> > > Karyn Ulriksen <kulriksen@publichost.com> writes:
> > >
> > > > What's the advantage/difference of GEIP+ over GEIP?
> > >
> > > Forget about using the GEIP+ for dot1q. Dunno about the GEIP.
> > > In any event, Not Ready For Prime Time.
> > >
> > > ---rob
> > >
>
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:42 EDT