My name is Naveed Babar, an Independent IT Expert and researcher. I received my Masters Degree an IT. I live in Peshawar, Khyber Pakhtunkhwa, Pakistan. Buzzwords in my world include: Info tech, Systems, Networks, public/private, identity, context, youth culture, social network sites, social media. I use this blog to express random thoughts about whatever I am thinking.

Monday, August 9, 2010

12 Things To Know When Troubleshooting Your Network

Introduction 

You work for a small organization where you wear many hats.  You are the network admin for PCs and your
routers and switches.  Anything that’s broken is YOUR problem and only your fault.  So, you walk into your of-
fice one morning, and you hear the phone ringing. You have an irate user on the other end telling you that the
“network is down.”  I always love those calls where a user, who has no networking background, tells you that
the network is down without any analysis tools or technical background.  After years of working in networking,
you probably know that if the network were really down, then the whole department would probably be calling
you and not just this one same user every morning.

This white paper approaches this situation with some of the common troubleshooting things that, you as a
network administrator, know or maybe should know.

Mental Preparation

The first and foremost concept for troubleshooting is not to panic. It might sound like the simplest thing, but
that is the reason why some new administrators make mistakes. Then, they dig themselves deeper because they
panic. A calm mind can identify the problem and approach the problem with a systematic method.  Sometimes,
if you have been working with the problem for a long time, it will not hurt to just walk away from the problem
and look at it after taking a break.  You’ll see that there are some very obvious things you missed.  I also like the
buddy system. You might want to have someone else look at the problem rather than you seeing or not seeing
the same issue - which was probably staring right at you.

Now, we know that in real life you can’t just walk away from a problem; your management might not appreci-
ate that.  You might be losing many millions of dollars while the network is down.  I can tell you that in certain
medical environments, you want to fix the problem fast since someone’s life might depend on that piece of
equipment that you are trying to fix.

Systematic Approach 

So, let’s visit our caller from the beginning of the paper.  We are going with an assumption that this user is
running the most common operating system available out there – some Windows platform. Common mistakes
that new administrators make are that they believe the user and assume that the network is really down.  I
would strongly recommend against that.  I am not asking you to argue with the user.  I would like to check out
the problem on my own.

Physical Layer Verification 

One of the first things you should do it to check the cable.  Do you have a green light at the back of your ma- 
chine?   You will find many times that the problem might just be a cabling problem. I have seen, in some rare 
cases, that inexperienced users just didn’t have the monitor plugged in so it was just a loose cable. Although, in 
my example, it could be the monitor cable, it could also be a network cable that is unplugged. 

Network Layer Verification 

If I see that the cable is plugged in, then I would ensure that there is an IP address assigned on the system. 
Now, with Windows systems, you can go to the command prompt and checking the IP address using the 
“ipconfig” command.  You want to make sure that you see that the IP address/subnet mask that are assigned 
are correct for that segment.    Keep in mind that the command to verify your IP address and subnet mask will 
vary, depending on the system you are working on.  For Windows systems, it’s ipconfig or ipconfig /all for more 
detailed information.  For Cisco routers, the command show interface will show you the IP address and subnet 
mask.  So, please, consult your documentation for the systems you are working on. 

Interviewing the User 

I have had many instances where, after all the work you do, you find out that someone had just moved this 
machine to this location from a different office, or floor, or segment, and it still has the IP address/subnet mask 
from a different part of the network. Maybe this system has both wired and wireless cards, but the IP address 
was incorrectly configured on the wireless card rather than on the wired connection.  Maybe it would’ve been 
wise to ask the user if something changed in the environment related to machines being moved. 

Reviewing the Logs: If You Have Them 

I hope that your company has a central change log procedure where all things are documented, and you can 
refer to the logs before you even approach the user.  Besides the company change logs, I hope you are also using
a central syslog facility that is collecting alerts from various systems in your organization.  
You can find many free syslog products on the internet for download. 

Knowing Your Company Policy 

It is also wise to ask the user the last time they were able to successfully connect to the network from this 
machine.    I had a user once who kept insisting that the company internet was down so he couldn’t do web 
browsing related to his job.  After further investigation, we found out that the websites this particular user was 
trying to visit were banned by the company and therefore prohibited.  It is a very good idea to know your own
company’s security and ethical policies.  I would recommend having the disclaimers in writing approved by your senior management. 

Isolating the Problem Using Tools 

It also makes sense to do a basic ping command to try to get a response to/from different systems on or off 
your network.  If you can reach the local systems but not the remote systems, then there is a possibility that your default gateway is down, or missing, or configured incorrectly.  It is also possible that your default gateway (router) doesn’t know how to route the packets to that particular destination, or maybe it’s possible that your company doesn’t allow ICMP protocol, which is what ping uses.  Again, it makes perfect sense to know your company policies. 

Sometimes, you can ping the remote systems with their IP address but not with their names.  This implies that 
there is no name resolution method available to resolve from the name to the IP address.It could also be that 
the name resolution method is resolving it to the wrong IP address. 

In Windows systems, you can also check to see if your TCP/IP stack is loaded correctly.  You can ping the local-host address (127.0.0.1), which can be done at the command prompt.  DNS does the translation from the name to IP address.  You can see what your DNS is configured for using the ipconfig/all command at the command 
prompt. 

You can also use the traceroute command (tracert on Windows Systems) to see if you can see where the 
packets are failing.  You can see a hop by hop packet flow using the tracert command.  Tracert command 
output is explained in various Microsoft and Cisco courses offered.

Here is a sample partial output of tracert command. 

         Microsoft Windows [Version 6.0.6001] 

         C:\Users\dt>tracert 10.0.0.1 

         Tracing route to test [10.0.0.1] 

         over a maximum of 30 hops: 

                  1        <1 ms <1 ms <1 ms . [192.168.2.1] 

                  2        6 ms     7 ms  5 ms  10.58.160.1 

                  3        6 ms     7 ms  7 ms  gig-2-0-nycmnyu-rtr2.nyc.rr.com 
         [24.29.98.189] 

         …….……. 

Have an Up-To-Date Network Diagram (Not an 
Outdated One) 

Let’s say you did all those things, and you see that all those things are configured correctly or not prohibited by 
your company policy.  What should you do next?  It’s always good to have a network topology map to consult 
to see where this particular system is supposed to be plugged in.    If there is nothing wrong on the PC, then the 
next logical approach will be to go to the switch where the user is plugged in.  In this white paper, we will only 
use Cisco Systems routers and switches in our examples. 

Checking the Switch 

Now that we are on the switch where the user is plugged in, and you have found the port number where the 
user is plugged in, let’s see how that particular port is configured. You can use the show run int gi 0/2 com- 
mand to see what is currently running for that particular interface (in this example gigabit 0/2 port) 

Here is a partial output from the show run int command. 

         Switch#show running-config int gi0/2 

         Building configuration... 

         Current configuration : 85 bytes 

         ! 

         interface Gigabit0/2 

         no ip address 

         switchport 

         switchport mode access 

         … 

         Switch# 

Another thing to check on the port would be to see if the port is in the right VLAN.  VLAN is a logical grouping 
of ports. 

VLAN assignment can be checked on Cisco switches using the Show vlan output. 

Here is a partial output from one of the switches showing the ports that are assigned to the two VLANS (namely 
VLAN1 and VLAN2).  Gigabit 0/2 – Gigabit 0/5 are assigned to VLAN1 and Gigabit 0/6 – Gigabit 0/12 are as- 
signed to VLAN2). 
Another thing to check on the port would be to see if the port is in the right VLAN.  VLAN is a logical grouping 
of ports. 

VLAN assignment can be checked on Cisco switches using the Show vlan output. 

Here is a partial output from one of the switches showing the ports that are assigned to the two VLANS (namely 
VLAN1 and VLAN2).  Gigabit 0/2 – Gigabit 0/5 are assigned to VLAN1 and Gigabit 0/6 – Gigabit 0/12 are as- 
signed to VLAN2). 

         Switch# show vlan 

        VLAN           Name             Status         Ports 

         --------- ----------- ---------- -------------------------------- 

         1             default          active         Gi0/2, Gi0/3, Gi0/4, Gi0/5 

         2             VLAN0002         active         Gi0/6, Gi0/7, Gi0/8, Gi0/9 

                                                                    Gi0/10, Gi0/11, Gi0/12 

         …….. ………. 

 All ports in the same VLAN form a logical grouping called a broadcast domain.  Broadcasts stay within these 
ports.  In our output example, Gi0/2 – Gi0/5 form one broadcast domain, and ports Gi0/6 – Gi0/12 forms the 
other broadcast domain.  All systems plugged in the same VLAN should have same subnet IP address.  
(It’s very important to have a very solid understanding of how subnetting works.   
This is also a typical problem in most environments.  
Now that you have verifed that the VLAN assignment is correct on the switch, you might want to ensure that 
there are no other restrictions on the switch port, such as port-security restricting that port to be used for 
only certain MAC addresses.  Typically, in organizations, it is common to see that companies implement MAC 
address-based security and then move machines around.  Perhaps the port is still only allowing the old machine 
and not this new system.  Remember, this could be a new machine assigned to the same user, or it could be 
that particular cubicle was used by someone else and now this is the new user.  Remember, both of these things 
could’ve been answered with a good conversation with the user or maybe looking at the log fle we discussed in
the beginning. Let’s say we verifed that the router is plugged in, and you are now connected to the router.  Does the router 
have a path to the destination where this particular person is trying to go?    Does the router know how to route
to that destination?   You can always use the show ip route command to see if the routes exist to the destina-
tion. However, in bigger environments, this might be very cumbersome since you might have hundreds or even 
thousands of routes. 
Here is a sample output of show ip route, looking only for 10.0.0.1 destination.  As you can see, you have 
a route to that destination.  In advanced routing classes, you learn what the other fun parameters mean.  The 
Global Knowledge BSCI class covers this in great detail.
Router# show ip route 10.0.0.1
Routing entry for 10.0.0.1/32
  Known via “isis”, distance 115, metric 20, type level-1
  Redistributing via isis
  Last update from 10.191.255.251 on Fddi1/0, 00:00:13 ago
  Routing Descriptor Blocks:
  * 10.22.22.2, from 10.191.255.247, via Serial2/3
  Route metric is 20, traffc share count is 1
……..………
Now that you know that you have a route to that destination, you can try pinging the destination. Remember, 
just because you have a path to some destination, that doesn’t mean that the remote site has connectivity with 
you.  How do you know that there is a return path available?   If both of these sites are in your offces, then 
you can ping from the remote site.   Also remember, it could be that your company doesn’t allow ping (or entire 
ICMP protocol).  Have you considered using something other than ping? You might try to do a telnet or even FTP 
to the remote site?  If the remote site is running the Telnet or the FTP daemon, then you will succeed.  If they are 
not running that, then maybe you need to fnd some other test application.
Most people think that since they couldn’t get in the remote site using Telnet or FTP, it means a failure.  As long 
as you even get a prompt for a password, that is a success.  You don’t have to be connected.  Getting the dialog 
box for password is a success.
If the user is trying to talk to systems that are outside of its broadcast domain, then there has to be a device 
that will take you outside this grouping of ports (VLANs).  How do we go out of our room?  Well, that’s simple, 
we just use the door.  So, what is the equivalent of the door in the networking world?  Your router is the door 
that takes you outside of your local network segment.  
Checking the Router
Let’s visit the router.  Is your router up?  Do you see the lights?   Again, the same basic question is whether it is 
powered up.  The port on the router that you are using as your default gateway for the PC should be plugged 
into the same VLAN.

No comments:

Comments

Search This Blog

Followers