Re: [nsp] Cisco 7206 as DSL concentrator: A BAD IDEA!

From: Siva Valliappan (svalliap@cisco.com)
Date: Mon Nov 13 2000 - 15:35:37 EST


Hi Salvatore,

   i am very sorry to hear about the issues you were experiencing. please
find some comments inline so that we can help you resolve the issues and
successfully deploy the 7200 in your network.

regards
.siva

>
> Hi all,
>
> I'm using a set of Cisco 7206 as DSL concentrator for 6 month now, and I can
> say only one thing: Cisco 7206 is not good as DSL concentrator.
>
> In my opinion Cisco is one of the most respected Company in the Network
> industry, but I can say that this time they maked a big flop.
>
> After 6 month of work in our DSL infrastructure, from the first day we noted
> a big problem in Cisco IOS, causing ATM to goes down without ANY reason, and
> after a lot of time we must use the same IOS revision, because EVERY version
> came out from Cisco after the first one as annoying Bugs that in the first
> version are not exhibited.
>
> A strange situation! below a list of the IOS that we tried, you can note
> that we tried every version come out from Cisco supporting PPPoE ( we need
> to support PPPoA and PPPoE).
>
> 1) 12.1.1aT1 - Is the first version that Cisco presented for use it 7200 and
> featuring PPPoE, we noted at the begin LOT of memory align errors, and later
> we encountered RANDOM ATM interface DOWN problems! From 1 problem in a week
> to 4 in a DAY. Strange things is that PPPoA session and PPPoE sessions
> worked well without any trouble (the only troubles came from ATM errors I
> think)

were there any log messages that accompanied the ATM interface going down?
did the interface go "down/down", "up/down", etc?

how did you recover from them? would you happen to have a "show tech"
that you obtained while the interface was down in this manner?

are you running ATM OAM on your PVCs by any chance? knowing the nature of
failure for the ATM interface can help us predict the cause of failure.
we can also suggest what additional information we need.

if you can provide the output of "show align" or a copy of the alignment
errors and their tracebacks, together with what feature set you were using
in 12.1(1a)T1, we can decode the alignment errors and figure out what
is happening.

please also provide the config of the router, so that we may be able to
make some configuration recommendations.

>
> 2) 12.1.2T - After 6 week of problems came out the next version of IOS, we
> tried it and immediate we noted that after a small amount of time it stopped
> to accept PPPoE sessions! All problems of 12.1.1aT1 are gone but a bigger
> one is here! Sticking on version 12.1.1aT1

can you please provide the config of the router? did you have a chance
to run any pppoe debugs while this problem was happening to see why we
stopped accepting PPPoE sessions?

>
> 3) 12.1.3aT series - This version will not authenticate any customer is you
> are using "aaa authorization network default group radius", it gives strange
> "AAA/FSM: Internal State is invalid".
>

you normally see this error message to indicate a possible AAA
misconfiguration. it is possible that 12.1(3a)T enforced something
more strictly, or some default in AAA changed, causing this behavior.
it is also likely that something else was the problem, and this error
message was a red herring.

the error message should have been accompanied by a Traceback. if you
still have this Traceback, and you can provide it to us, together with
the exact image [we need to know version and feature set], we can decode
the traceback and determine the cause of the problem. did you have a
chance to run any AAA debugs while this problem was happening?

> 4) 12.1.3aXX - When starting with this IOS all our customers using routers
> cannon browsing any more, ping traceroutes are going quite well. Another
> STRANGE thing!

sounds like a Path MTU problem. were only your PPPoE users affected, while
the PPPoA users were fine? if it's a Path MTU problem, it should be
affecting the users regardless of IOS version that you are running. the
problem derives from the fact that PPPoE has a max IP datagram size of 1492
[you lose 8 bytes for the PPPoE header]. if the 2 end devices negotiate
a max TCP size of 1500, and if there is ICMP filtering somewhere along the
path, the web server that is sending the 1500 byte packets, will not get
the ICMP MTU too big error messages that the router will generate.

>
> Can some one help us?? There is out another IOS version to try? Any config
> help? If you need other information send me a e-mail.
>

we have been doing an incredible amount of software work to scale the
xDSL aggregation features for the 7200. we hope to release 12.1(5)T soon
with support for up to 4000 sessions :) we have also been testing the
enhanced scalability images with some of our customers.

12.1(5)T is not out yet, but i can site publish 12.1(4.4)T3 to you. this
image is TAC supported [it is an interim image]. most of the scalability
work and bug fixes are in this code.

i would urge you to open a TAC case, so that TAC can work with you in
deploying this image, and if anything goes wrong, work with you in
gathering the information we need to troubleshoot the problem.

please inform TAC that you have been working with me, and to use me as
a point of escalation.

 
> Thanks
> Salvo
>



This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:20 EDT