Q: I
got the key, but I couldn't open the door, is there combination code?
A: No, there is no combination code. You
might want to try the lower keyhole.
Q: What is time slot for lab
exercises?
A: Because we have limited number of routers, each student is assigned several
time slots each week for doing lab exercises. During your assigned time slots,
you have exclusive use of the assigned routers. Each time slot is four-hour
long. There are two sets of routers, called Rack I and Rack II. So, each week
has 2 (racks) * 7 (days) * 6 (slots/day) = 84 time slots.
Q: What are the passwords of Linux machines
and routers?
A: The password of Linux machines is on the lab whiteboard. Routers don't have
passwords. DO NOT CHANGE ANY PASSWORD. In
case when you configure routers and are asked to "Enter router
password:" or "Enter router secret:", just do <CTRL-C>
to abort the configuration dialogue. You will come to command line prompt. You
are supposed to configure the router at command line.
Q: How will our submissions be graded?
A: Most points are given to the trace files and the reports. The trace files
can show whether you did the exercises correctly. Some questions ask you about
the phenomena of the exercises, you should explain why they happen, not just
answer yes or no. You can also write your thoughts or comments about the
exercises in the end of the report to help us improve them.
Q: I can't get minicom to work,
how should I set the parameters?
A: /dev/ttyS0, 9600 baud rate, 8N1, No flow control.
Q: Can I download router's configuration file
to PC?
A: Yes, you can use tftp to download and upload router's running-config.
Read the man pages and Cisco documentation
to figure out how to do it.
Q: I can't find tftp server and client in
Linux machines?
A: The tftp server and client may not be installed in Linux. Get the Linux CDs
and install tftp-*.rpm. The server should be called in.tftpd
and the client tftp.
Q: Do I need to clean up the configuration
information on the routers and the Linux machines? Can you tell me what should
I do?
A: Yes, you should clean up all your configuration of both the Cisco routers
and the Linux machines. To erase the configuration on Zebra, you may delete
the configuration files (zebra.conf, ospfd.conf, bgpd.conf) under
<where-you-install-zebra>/etc. To erase the configuration
on Cisco routers, do a "clean startup-config".
Q: I've set the IP address on the
corresponding interfaces, but routers can't ping each other. Do you have any
suggestion?
A: I assume you've made the correct IP address assignment. First, you should
make sure the the cable connection is Ok. Second, the Ethernet interface on
the router shall be UP (you should figure out the Cisco command to bring up
the interface). If the problem is still not resolved, send email to TA.
Q: How do I configure F such that it belongs
to different AS?
A: Multiple instances on bgpd.
Q: I think the Ethernet card on F is not
working, I couldn't access to the Internet even if I set the IP address,
gateway and DNS correctly.
A: There are 2 Ethernet cards on F, the upper one (eth1) is supposed to
connect to the external LAN. The problem is due to the fact that you configure
eth1 but plug the cable to the lower NIC (eth0).
Q: I'm trying to configure A, C and F into
one subnet...
A: You don't need to do that. All nodes in a subnet must be able to communicate with each
other directly (e.g. they are all connected to the same Ethernet segment).
Q: Do I need to configure virtual interface
on C, D, E?
A: If you want to keep Internet access on these machines, you should.
Q: Router A and B do not exchange BGP info,
why?
A: This is a common problem that happened to many students, it is very likely
that something is wrong in the configuration of Router A and B. You might want
to do "no synchronization" in A and B. And setting the
cluster-id may help. Here
is a pointer to the Cisco page on troubleshooting when BGP routes are not
advertised.
Q: I have problem in setting up multiple
instances on F, do I need to use view?
A: Yes. The current Zebra implementation seems not supporting BGP multiple
instance perfectly. But If you manage to set it without using view, please let
me know...
Q: How do I know whether the route reflector
is working or not?
A: One good way to debug the configuration is to collect traces on the
routers. On zebra machines, you need to enable debug to collect the trace (the
zebra manual doesn't say this). By analyzing the trace, you will get many
hints on what happened and what might go wrong.
Q: Can you explain why we need "no
synchronization"?
A: When synchronization is enabled, the route reflector will not advertise the
routes it learns from its clients to other BGP peers (outside that cluster) if
the route can't be validated in IGP. We need to do this since F is not running
OSPF and the route advertised by F is usually non-existent.
Q: What is the difference between routing
table and BGP routing
table?
A: Routing tables store the routes that routers use to forward packets. BGP routing
tables store only the BGP routes that are dynamically learned from neighbors or
statically configured. Similarly, OSPF routing
tables store only the OSPF routes.
Routers will merge static routes, OSPF routes, and BGP routes into the routing
tables.
Q: How to get the trace and BGP routing
messages on zebra?
A: Put the following commands in bgpd.conf:
debug bgp updates
log file
bgp.trace
dump bgp all bgp.packets
The trace file (bgp.trace) is in text format, while the dump file
(bgp.packets) is in binary format which is called MRT format. There is a tool in the MRTd package
(http://www.mrtd.net/) that can view the dump file. You can find an executable
of the tool: http://catarina.usc.edu/~difa/cs558/route_btoa
This tool will generate output like:
% ./route_btoa bgp.packets
You should submit the original binary dump files. Do not convert it to
text format.
TIME: 1092084092 [08/09/04 13:41:32]
TYPE: BGP4MP/MESSAGE/Update
FROM: 198.32.16.49 AS4
TO: 198.32.16.54 AS65530
ORIGIN: IGP
ASPATH: 4
NEXT_HOP: 198.32.16.49
MULTI_EXIT_DISC: 0
ANNOUNCE
192.228.79.0/24
Some suggestions:
1. Make sure you have a clean system, reinstall the system if you think it is
necessary;
2. Make sure the network connection is working, also you should check it from time
to time;
3. If C, D, and E complain that they can't find the route to the network that F
advertises, try adding a static route on C, D, and E, telling them that it can be
reached through F.
Q: There are no enough transceivers in the
lab. How can we setup the topology?
A: Instead of connecting two machines via a direct link, you can connect
multiple machines via the hub. The physical topology may not resemble to the
one shown in the exercise, but the logical topology should be the same. Hint: make sure the target and the sinkhole use
dedicated links, so when you run tcpdump on these machines, it won't capture the traffic not belonging to them.
Q: Do we have to put all machines and routers in the same
AS?
A: Yes.
Q: When I trigger the blackhole or sinkhole
routes, they don't propagate to other routers. Why?
A: Remember that BGP
requires the I-BGP peerings in full mesh. You can setup a full mesh
peering topology, or use route reflection you learned in exercise 1.
Depending on your configuration, you may also need to run OSPF.
Q: Do we need to run zebra in machines?
A: No, you can configure static routes in machines so that they know how to reach each other.
Q: What you mean by sending back-to-back UDP
packets?
A: It means you should send UDP packets as fast as possible.
Q: How can I know the TOS octets of packets
are set correctly?
A: You can use tcpdump to check the trace files. If packets have the tos octet
set, tcpdump will show something like [tos 0xAB] where AB is the
value of tos octet. Remember that precedence field is the first 3-bit of tos
octet.
Q: My submission tarball is huge, how can I
submit it?
A: You can use lab machines or public email account (e.g. Yahoo, Gmail) to email
them. If you use lab machines, remember to set the email's
"Reply-To" to your regular email address where the confirmation mail
will be sent.