Mailing List Archive

Tue, Mar 12, 2002 at 02:00:51PM -0500, Robert Klingsten:
> Hello --
>
> I have read through all the list archives, and also the faq. But
> I am still having trouble with a Catalyst switch. Following the
> diags mentioned in the FAQ, the command:
>
> cat5rancid -d ummu-cis6513-mr.ummu.umich.edu
>
> results in:
>
> HIT COMMAND:ummu-cis6513-mr> (enable) show port ifindex
> In ShowPortIfindex: ummu-cis6513-mr> (enable) show port ifindex
> HIT COMMAND:ummu-cis6513-mr> (enable) write term
> In WriteTerm: ummu-cis6513-mr> (enable) write term
> ummu-cis6513-mr.ummu.umich.edu: End of run not found
> ummu-cis6513-mr.ummu.umich.edu: End of run not found
> end

could you try the following

setenv NOPIPE YES
cat5rancid -d ummu-cis6513-mr.ummu.umich.edu

you will have a ummu-cis6513-mr.ummu.umich.edu.raw file. check this to
make sure that the full config was actually collected. also check the
ummu-cis6513-mr.ummu.umich.edu.new file to see how much of the .raw file
was completed.

i'd guess that either the expect process got stuck or something appeared
in the config that confused clogin.

>
> Any more help from people on this would be greatly appreciated! I have
> Linux kernel 2.4.18, expect 5.32.2, tcl 8.3.3-65, Catalyst 6.3(5)...
>
> thanks for any help!
>
> Rob Klingsten
> CAEN Network Systems
> University of Michigan College of Engineering

I tried what you suggested. the .new file has the complete config, and the
.raw file has the complete config and ends with:

.
.
.
#module 15 : 1-port Multilayer Switch Feature Card
!
#module 16 empty
end
ummu-cis6513-mr> (enable)Connection to ummu-cis6513-mr.ummu.umich.edu closed
by remote host.
Connection to ummu-cis6513-mr.ummu.umich.edu closed.


It looks ok to me... is expect not matching the "Connection to <blah>
closed" message(s)?

Thanks --

Rob

-----Original Message-----
From: john heasley [mailto:heas@shrubbery.net]
Sent: Tuesday, March 12, 2002 3:16 PM
To: Robert Klingsten
Cc: rancid-discuss at shrubbery.net
Subject: Re: Newbie: Problems with Catalyst


Tue, Mar 12, 2002 at 02:00:51PM -0500, Robert Klingsten:
> Hello --
>
> I have read through all the list archives, and also the faq. But I am
> still having trouble with a Catalyst switch. Following the diags
> mentioned in the FAQ, the command:
>
> cat5rancid -d ummu-cis6513-mr.ummu.umich.edu
>
> results in:
>
> HIT COMMAND:ummu-cis6513-mr> (enable) show port ifindex
> In ShowPortIfindex: ummu-cis6513-mr> (enable) show port ifindex
> HIT COMMAND:ummu-cis6513-mr> (enable) write term
> In WriteTerm: ummu-cis6513-mr> (enable) write term
> ummu-cis6513-mr.ummu.umich.edu: End of run not found
> ummu-cis6513-mr.ummu.umich.edu: End of run not found
> end

could you try the following

setenv NOPIPE YES
cat5rancid -d ummu-cis6513-mr.ummu.umich.edu

you will have a ummu-cis6513-mr.ummu.umich.edu.raw file. check this to make
sure that the full config was actually collected. also check the
ummu-cis6513-mr.ummu.umich.edu.new file to see how much of the .raw file was
completed.

i'd guess that either the expect process got stuck or something appeared in
the config that confused clogin.

>
> Any more help from people on this would be greatly appreciated! I
> have Linux kernel 2.4.18, expect 5.32.2, tcl 8.3.3-65, Catalyst
> 6.3(5)...
>
> thanks for any help!
>
> Rob Klingsten
> CAEN Network Systems
> University of Michigan College of Engineering

On Tue, Mar 12, 2002 at 03:32:27PM -0500, Robert Klingsten wrote:
> ummu-cis6513-mr> (enable)Connection to ummu-cis6513-mr.ummu.umich.edu closed
> by remote host.
> Connection to ummu-cis6513-mr.ummu.umich.edu closed.

Looks like it could be another case of a router being too eager to
drop the tcp session and not echoing back the 'quit' rancid sends
it first.

Rancid sends a 'quit' and then the router should echo that back &
then close the tcp session (it should wait for the output to drain
before doing the final tcp close). Some routers have bugs and
don't do this.

I know we had to do a fair amount of expect goo to work around this
for the extreme (in xrancid).
--asp

> I tried what you suggested. the .new file has the complete config, and the
> .raw file has the complete config and ends with:
> ummu-cis6513-mr> (enable)Connection to ummu-cis6513-mr.ummu.umich.edu closed
> by remote host.
> Connection to ummu-cis6513-mr.ummu.umich.edu closed.
>
> It looks ok to me... is expect not matching the "Connection to <blah>
> closed" message(s)?

Are you using SSH to talk to the switch? I sent in a patch for this
problem a month or two ago. Take a peek in the mailing list archives.

After thinking about it, and the various other platforms where it
happens, I think a better solution would be to invoke a harmless com-
mand (for CatOS, something like "cd end-of-run:") and pick up on the
"end-of-run" token instead of hoping some combination of Cisco OS,
client SSH, and host TCP/IP stack will bother echoing the "exit" com-
mand back to us before the connection gets torn down.

Of course, on devices that have an "echo"-like command, this is
even more elegant as we don't generate an (ignored) syntax error.

Terry Kennedy http://www.tmk.com
terry at tmk.com New York, NY USA