Monday, April 15, 2013

Network connectivity - the bane of all computer users...

Network connectivity - the bane of all computer users...

If I am faced with a machine with a broken network connection, I always start with a few simple tests.  First see whether the machine has an IP address, netmask and default route.  If the basics do not work, nothing else will, but it is amazing how many people do not think about that!

If the basics are fine, try to do a ping of another machine - that will test most things, including the Domain Name Server.  Finally, I use Telnet to verify that the destination service is working.

Doing these tests manually, will provide you with helpful error messages.   The Linux error messages always tell you *exactly* what the problem is.  If you don't understand them, then google these messages for more information.  Another unfortunate soul likely had the same problem already.  If you use a GUI utility, then you are insulated from the error messages and you are reduced to aimlessly poking around in the dark.

If those do not work, try to use the Dynamic Host Configuration Protocol (DHCP) to fix things.


Display the device name, IP address and netmask:
# ifconfig

 em1       Link encap:Ethernet  HWaddr 00:21:CC:5E:98:D9
          inet addr:  Bcast:  Mask:

In the olden days, the ethernet ports were always named eth0, eth1 and so on, but nowadays it seems that it can be anything at all, for example em1 or whatever the device file system chose at first startup.

You can also use ifconfig to set the IP address of an embedded system:
# ifconfig em1 netmask up


Display the routing table and make sure that there is a default gateway defined:
# route

Kernel IP routing table
Destination     Gateway         Genmask         Flags Metric Ref    Use Iface
default         UG    0      0        0 em1      *        U     1      0        0 em1


You can use dhclient to talk to a DHCP server to reset the machine address and route (make sure it is stopped before running it again):
# killall dhclient
# dhclient eth0

If there is no DHCP server, if for example you are running an embedded system with static addresses, then you got to configure the above manually and in this case, you should disable automatic wizards such as NetworkManager, since it will cause you periodic grief by ever so helpfully changing your static setup for you!


If all works so far, try to ping something.  Yahoo always answers to pings:
# ping

PING ( 56(84) bytes of data.

Some network administrators configure their routers to filter out ping requests.  This is rather unhelpful, because ping is a valuable test tool, but Telnet will accomplish much the same thing.


Ping also shows whether the DNS is working.  If not, check the nameserver definition in /etc/resolve.conf.  You can always use the Google DNS at if you don't have a better one.

You can also do a DNS test with nslookup or dig:
# dig

; <<>> DiG 9.8.4-P1-RedHat-9.8.4-3.P1.fc16 <<>>
;; global options: +cmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 10032
;; flags: qr rd ra; QUERY: 1, ANSWER: 5, AUTHORITY: 0, ADDITIONAL: 0

;            IN    A

;; ANSWER SECTION:        0    IN    CNAME    0    IN    CNAME    0    IN    CNAME 0 IN    CNAME 42    IN    A

;; Query time: 12 msec
;; WHEN: Mon Apr 15 12:30:45 2013
;; MSG SIZE  rcvd: 151


Finally, try to connect to the destination server using Telnet and see what it says:
# telnet 80

Connected to
Escape character is '^]'.

You should get the web server banner and you could even GET the home page manually:

get index.html
<BODY BGCOLOR="white" FGCOLOR="black">
<FONT FACE="Helvetica,Arial"><B>

<!-- default "Redirect" response (302) -->
Connection closed by foreign host.

Happy debugging!

No comments:

Post a Comment

On topic comments are welcome. Junk will be deleted.