[prev in list] [next in list] [prev in thread] [next in thread] 

List:       cisco-nsp
Subject:    RE: [nsp] dCEF and IOS
From:       George Robbins <grr () shandakor ! tharsis ! com>
Date:       2000-04-25 6:07:06
[Download RAW message or body]

One of the excuses for dCEF failures is not enough memory on the VIP's
(the 12.x VIP images that are downloaded into the VIP's are larger than
the corresponding 11.x images), but it's less clear that adding memory
will really solve the problem.

I've had to back off a couple of our CEF'd routers to flow switching,
just becuase there was no way to keep them reliable and the problems
don't fall into a simple category that our NOC can identify and resolve
without escalation.

High-risk situations seem to be a lot of sub-interfaces, channelized
t3/e3 cards or lots of secondary addresses on an interface.  Clearly
there are many bugs left in the CEF world, I wonder if Cisco has a
team working on it, or it's assumed to be mostly OK and cases just
run thru the mill with no special priority.

						George

> From: "Rubens Kuhl Jr." <rkuhljr@uol.com.br>
> To: "Rich Sena" <ras@poppa.thick.net>
> Cc: <cisco-nsp@puck.nether.net>
> Subject: RE: [nsp] dCEF and IOS
> Date: Mon, 24 Apr 2000 14:27:10 -0300
> In-Reply-To: <Pine.LNX.4.10.10004241240170.24001-100000@poppa.thick.net>
> Resent-From: cisco-nsp@puck.nether.net
> X-Mailing-List: <cisco-nsp@puck.nether.net> archive/latest/2721
> X-Loop: cisco-nsp@puck.nether.net
> Precedence: list
> Resent-Sender: cisco-nsp-request@puck.nether.net
> 
> 
> By memory issue you mean memory failure or lack of memory (either SRAM or
> DRAM) ?
> 
> 
> Rubens Kuhl Jr.
> 
> 
> > -----Original Message-----
> > From: Rich Sena [mailto:ras@poppa.thick.net]
> > Sent: Monday, April 24, 2000 1:43 PM
> > To: Martin, Christian
> > Cc: 'cisco-nsp@puck.nether.net'
> > Subject: Re: [nsp] dCEF and IOS
> >
> >
> > I'm seeing the exact same probs - I have been directed by TAC that it is a
> > memory issue on the VIP2 50's - I have been hesitatnt to accept this -
> > however I see no other real explaination - any light anyone can shed on it
> > is apreciated...
> >
> > --
> > I am nothing if not net-Q! - ras@poppa.thick.net
> >
> > On Mon, 24 Apr 2000, Martin, Christian wrote:
> >
> > > Hello, folks.
> > >
> > > Can someone make a recommendation on a software image that runs dCEF
> > > cleanly?  I am seeing FIB corruption due to bad Adj data.  I
> > am seeing high
> > > numbers for IPC Seat Manager, cpu-wise.  The corruption
> > manifests itself as
> > > an inability to forward packets out certain interfaces, in no particular
> > > pattern.  Packets with options get thru, so processs-switching
> > is working.
> > > Clearing cef adj, or clear cef line works sometimes -
> > sometimes, I have to
> > > disable and re-enable cef.  Sometimes doing this causes the
> > router to crash.
> > > I am on all 2.02 VIP 2-50s.  Doesn't appear to be a VIP issue,
> > no crashes,
> > > IPC message getting to VIPs, if-con to VIP shows good IPC status, etc.
> > > 12.0.10S is available, I believe, but tehre are some bad caveats.  I've
> > > tried 12.08 and 12.09S.  11.1.26CC1 runs cleanly, but we lose ATM SNMP
> > > support and ATM subinterface ifIndex support, so I _really_
> > don't want to go
> > > back to 11.1.
> > >
> > > TIA,
> > > Chris
> > >
> > > PS
> > >
> > > I also have a very large ARP cache (15,000+), which may be
> > causing the IPC
> > > Seat churn.
> > >
> >
> 
> 

[prev in list] [next in list] [prev in thread] [next in thread]