Debian: Network Configuration

Network configuration in Linux Debian. This topic is intended to become a cheat sheet for a system administrator on "how to manage the network in Debian". It describes the basic principles and actions for managing network settings on this family of operating systems.


Network interfaces

Debian names network cards by the name eth and a number. Virtual interfaces (loopback, vpn, etc.) Debian names separately, likewise by a name (for example, tun, tap) followed by a number.

For example, if you have 2 network cards in your computer, then by default they will be named eth0 and eth1.

Note that this operating system permanently binds the MAC address of the physical network interface (network card) to the ethX number. So, if you had a network card in the computer named eth0 and you replaced it, then after booting following the replacement you will not see the expected eth0, but will instead see eth1.

As a particular but illustrative case: a server with two network interfaces eth0 and eth1. One of the network cards burns out and the administrator has to replace it. After starting the server, he never gets a response to a ping on that network interface. Logging into the console - what does he see? He sees the old second network card named eth1, and the new one, but not named eth0 as expected, instead named eth2! This is because Debian has bound a specific interface with a specific MAC address to eth0.

You can view the bound cards and their MAC addresses at the path:

/etc/udev/rules.d/70-persistent-net.rules

By editing this file you can remove old network cards from Debian's list.


IP addresses

Persistent settings are stored in the file /etc/network/interfaces

Here is an example file:


In this file the IPv4 address 192.168.0.1 is defined on the network card eth0. The default gateway 192.168.0.250 is also defined. The dns-nameservers option is left in case the file /etc/resolv.conf lacks DNS server information.

Note that modern Debian does not need the network and broadcast lines, computing them perfectly well on its own from the IP address and mask. However, if you want to override these values, you will need to specify such options.

And here is an example of this file for 2 network cards:



You can change IP address values on the fly using the ifconfig utility:


This command will change the IPv4 address of the network card eth0 to 192.168.0.33 with a /24 mask.


If network settings need to be obtained via DHCP - specify this in the interfaces file, for example:



Specifying the DNS server

DNS servers are specified (no more than three) in the file:

/etc/resolv.conf

For example, here is an example of such a file:


After changing the values in the file, they take effect immediately.


Machine name and domain

You can view or change the machine name on the fly with the hostname command:


The domain is specified in the file:

/etc/resolv.conf

for example, here is an example of a resolv.conf file:


The machine name can be changed permanently in the file:

/etc/hostname

But remember that after this you also need to change the hostname in the file

/etc/hosts

So that the values in it correspond to the new hostname (things like 127.0.0.1 are specified there, for example).


Routing

The static routing scheme in Debian is traditional: static routes and a default gateway.

Routing can be managed on the fly through the route utility:


This command will add a route to the network 10.0.3.0:255.255.255.0 via the router with address 192.168.0.251.


And this command, conversely, will remove the route 10.0.3.0/24.

The default router is specified as 0.0.0.0/32 or default, for example:


This command will add the default gateway 192.168.0.250.

You can view the current routes with the command:



Static routes are specified permanently in the file

/etc/network/interfaces

For example, let's take the following file:


From this file we can see that:

a) The default gateway is 192.168.0.250

b) When the interface comes up, routes to networks 10.0.0.0/24, 10.0.1.0/24 and 10.0.2.0/24 will be added via, respectively, 192.168.0.251, 192.168.0.252 and 192.168.0.253.


Router

By default the system does not route passing packets. To enable this (turn our server into a router) you need to set the kernel variable


to the value "1".

For example:


To make this value persist after a reboot - add it to the file

/etc/sysctl.conf

with the following line:



Network aliases (multiple IP addresses on one interface)

It is not uncommon to encounter a situation where one network card (or simply one network interface) needs to be assigned not one, but several IP addresses. In that case:

a) A primary address must be assigned to the ethX interface
b) Aliases (additional IP addresses) are assigned as ethX:Y, where Y is the alias number

For example, this is how we would assign 2 more IP addresses to the eth0 interface:


These interfaces are specified permanently in the same file as all the others:

/etc/network/interfaces

Let's give an example of such a file:


Here we can see that we specified one primary address and 3 aliases. Note that the gateway or dns-nameservers options cannot be applied to aliases.


VLANs

First, make sure the vlan package is installed:


If not, install it:


So that VLANs are set up on reboot, they need to be listed in the file

/etc/network/interfaces

For example, here is an example of such a file:


Here we see the declaration of 3 VLANs. Both the vlan333 notation and the eth0.999-style notation are correct. The difference will only be in their naming as interfaces. I prefer the eth.XXX form, which immediately tells me which interface the VLAN sits on.

The vlan_raw_device option was also added, indicating to the system which network card this VLAN will be tagged on.

VLANs can be managed at runtime through the vconfig utility. For example:


This command will add VLAN=777 to the eth0 interface.


And this command will remove the VLAN vlan777.


Network bridges

You can read about what this is on the internet. To set one up we need bridge-utils installed. Let's check for it:


If not present, install it:


Bridges are specified in the file

/etc/network/interfaces

For example, like this:


Here you can see that we do not configure the network cards themselves (we do not assign them an IP address, we do not let them get settings via DHCP). This must be done this way!

We name the network bridge br0. It is on this bridge that we specify all the necessary settings and the bridge_ports option, which shows which network interfaces to include in this bridge.

On the fly, bridges are managed with the brctl utility. For example, this is how you can create a network bridge from 2 network cards (the cards must not be configured and must be in the down state):


This sequence of commands:
a) Creates the virtual switch br0
b) Adds the network interfaces eth0 and eth1 to it
c) "Brings up" the bridge (up), i.e. enables it

After this, the bridge can be given network settings (or not - in that case these network cards will work as an unmanaged switch, and it will not be possible to connect to them from outside via these connections).


Other examples:


This command will remove the eth1 interface from the virtual switch (bridge).


And this command will remove the br0 bridge entirely.


Link aggregation (Bonding)

Interface aggregation is a way to increase data transfer speed and/or provide fault tolerance by using two or more physical network cards. It works so that several network cards act as one virtual one (not to be confused with a bridge!), and depending on the operating mode, these network cards either increase throughput (by a factor equal to the number of network cards combined), or provide fault tolerance (i.e. one card works, and if it dies, the second one starts working), or both.

This requires the ifenslave-2.6 package to be installed. If it isn't installed, install it:


The configuration is specified in the file

/etc/network/interfaces

For example, like this:


Here you can see that the physical cards are not configured - they are not assigned static IP addresses and are not allowed to use DHCP to get settings. Instead, they participate in creating the bond0 interface, which itself receives the network settings.

ATTENTION! On the other end of the network cables, the network cards must be configured identically (i.e. operate in the same mode). If these are hardware switches or routers, you will have to pick the operating mode - whichever is supported by the device on the other end. If it is a server, it is best to use:

a) the same network cards (if possible)
b) the same OS (this will reduce the likelihood of aggregation failures)


Automatically running scripts when an interface comes up/down

4 directories serve this purpose:

/etc/network/if-pre-up.d
/etc/network/if-up.d
/etc/network/if-down.d
/etc/network/if-post-down.d

Accordingly, the if-pre-up.d directory holds scripts that run before the interface comes up; if-up.d holds scripts that run after the interface comes up; if-down.d holds scripts that run before the interface goes down, and if-post-down.d holds scripts that run after it goes down.

Among other things, the scripts being called can use special variables from the calling process's environment, for example $IFACE (the interface name) or $MODE (the operation). For more details - see the man pages.

See also

• [[b6537]]

See also