layout-3 layout-6 layout-1 layout-top
NVIDIA Docs Hubchevron_right NVIDIA Networkingchevron_right Networking Softwarechevron_right Switch Softwarechevron_right

Cumulus Linux

layout-top layout-6 layout-1

If you are using the current version of Cumulus Linux, the content on this page may not be up to date. The current version of the documentation is available here. If you are redirected to the main page of the user guide, then this page may have been renamed; please search for it there.

Interface Configuration and Management

Cumulus Linux uses ifupdown2 to manage network interfaces, which is a new implementation of the Debian network interface manager ifupdown.

Bring an Interface Up or Down

An interface status can be in an:

  • Administrative state, where you configure the interface to be up or down.
  • Operational state, which reflects the current operational status of an interface.

To configure and bring an interface up administratively, use the nv set interface command:

cumulus@switch:~$ nv set interface swp1
cumulus@switch:~$ nv config apply

After you bring up an interface, you can bring it down administratively by changing the link state to down:

cumulus@switch:~$ nv set interface swp1 link state down
cumulus@switch:~$ nv config apply

To bring the interface back up, change the link state back to up:

cumulus@switch:~$ nv set interface swp1 link state up
cumulus@switch:~$ nv config apply

To remove an interface from the configuration entirely, use the nv unset interface command:

cumulus@switch:~$ nv unset interface swp1
cumulus@switch:~$ nv config apply

To configure and bring an interface up administratively, edit the /etc/network/interfaces file to add the interface stanza, then run the ifreload -a command:

cumulus@switch:~$ sudo nano /etc/network/interfaces
auto lo
iface lo inet loopback
    address 10.10.10.1/32
auto mgmt
iface mgmt
    address 127.0.0.1/8
    address ::1/128
    vrf-table auto
auto eth0
iface eth0 inet dhcp
    ip-forward off
    ip6-forward off
    vrf mgmt
auto swp1
iface swp1
...

To bring an interface down administratively after you configure it, add link-down yes to the interface stanza in the /etc/network/interfaces file, then run ifreload -a:

auto swp1
iface swp1
 link-down yes

If you configure an interface in the /etc/network/interfaces file, you can bring it down administratively with the ifdown swp1 command, then bring the interface back up with the ifup swp1 command. These changes do not persist after a reboot. After a reboot, the configuration present in /etc/network/interfaces takes effect.

By default, the ifupdown and ifup command is quiet. Use the verbose option (-v) to show commands as they execute when you bring an interface down or up.

To remove an interface from the configuration entirely, remove the interface stanza from the /etc/network/interfaces file, then run the ifreload -a command.

For additional information on interface administrative state and physical state, refer to this knowledge base article.

Interface Classes

ifupdown2 enables you to group interfaces into separate classes. A class is a user-defined label that groups interfaces that share a common function (such as uplink, downlink or compute). You specify classes in the /etc/network/interfaces file.

The most common class is auto, which you configure like this:

auto swp1
iface swp1

You can add other classes using the allow prefix. For example, if you have multiple interfaces used for uplinks, you can define a class called uplinks:

auto swp1
allow-uplink swp1
iface swp1 inet static
    address 10.1.1.1/31

auto swp2
allow-uplink swp2
iface swp2 inet static
    address 10.1.1.3/31

This allows you to perform operations on only these interfaces using the --allow=uplinks option. You can still use the -a options because these interfaces are also in the auto class:

cumulus@switch:~$ sudo ifup --allow=uplinks
cumulus@switch:~$ sudo ifreload -a

If you are using Management VRF, you can use the special interface class called mgmt and put the management interface into that class. The management VRF must have an IPv6 address in addition to an IPv4 address to work correctly.

allow-mgmt eth0
iface eth0 inet dhcp
    vrf mgmt

allow-mgmt mgmt
iface mgmt
    address 127.0.0.1/8
    address ::1/128
    vrf-table auto

All ifupdown2 commands (ifup, ifdown, ifquery, ifreload) can take a class. Include the --allow=<class> option when you run the command. For example, to reload the configuration for the management interface described above, run:

cumulus@switch:~$ sudo ifreload --allow=mgmt

Use the -a option to bring up or down all interfaces with the common auto class in the /etc/network/interfaces file.

To administratively bring up all interfaces marked auto, run:

cumulus@switch:~$ sudo ifup -a

To administratively bring down all interfaces marked auto, run:

cumulus@switch:~$ sudo ifdown -a

To reload all network interfaces marked auto, use the ifreload command. This command is equivalent to running ifdown then ifup; however, ifreload skips unchanged configurations:

cumulus@switch:~$ sudo ifreload -a

Cumulus Linux checks syntax by default. As a precaution, apply configurations only if the syntax check passes. Use the following compound command:

cumulus@switch:~$ sudo bash -c "ifreload -s -a && ifreload -a"

For more information, see the individual man pages for ifup(8), ifdown(8), ifreload(8).

Loopback Interface

Cumulus Linux has a preconfigured loopback interface. When the switch boots up, the loopback interface called lo is up and assigned an IP address of 127.0.0.1.

The loopback interface lo must always exist on the switch and must always be up.

To configure an IP address for the loopback interface:

cumulus@switch:~$ nv set interface lo ip address 10.10.10.1
cumulus@switch:~$ nv config apply

Edit the /etc/network/interfaces file to add an address line:

auto lo
iface lo inet loopback
    address 10.10.10.1

  • If the IP address has no subnet mask, it automatically becomes a /32 IP address. For example, 10.10.10.1 is 10.10.10.1/32.
  • You can configure multiple IP addresses for the loopback interface.

Subinterfaces

On Linux, an interface is a network device that can be either physical, (for example, swp1) or virtual (for example, vlan100). A VLAN subinterface is a VLAN device on an interface, and the VLAN ID appends to the parent interface using dot (.) VLAN notation. For example, a VLAN with ID 100 that is a subinterface of swp1 is swp1.100. The dot VLAN notation for a VLAN device name is a standard way to specify a VLAN device on Linux.

A VLAN subinterface only receives traffic tagged for that VLAN; therefore, swp1.100 only receives packets that have a VLAN 100 tag on switch port swp1. Any packets that transmit from swp1.100 have a VLAN 100 tag.

The following example configures a routed subinterface on swp1 in VLAN 100:

cumulus@switch:~$ nv set interface swp1.100 ip address 192.168.100.1/24
cumulus@switch:~$ nv config apply

Edit the /etc/network/interfaces file, then run ifreload -a:

cumulus@switch:~$ sudo nano /etc/network/interfaces
...
auto swp1.100
iface swp1.100
 address 192.168.100.1/24

cumulus@switch:~$ sudo ifreload -a

  • If you are using a VLAN subinterface, do not add that VLAN under the bridge stanza.
  • You cannot use NVUE commands to create a routed subinterface for VLAN 1.

Interface IP Addresses

You can specify both IPv4 and IPv6 addresses for the same interface.

For IPv6 addresses:

  • You can create or modify the IP address for an interface using either :: or 0:0:0 notation. For example, both 2620:149:43:c109:0:0:0:5 and 2001:DB8::1/126 are valid.
  • Cumulus Linux assigns the IPv6 address with all zeroes in the interface identifier (2001:DB8::/126) for each subnet; connected hosts cannot use this address.

The following example commands configure three IP addresses for swp1; two IPv4 addresses and one IPv6 address.

cumulus@switch:~$ nv set interface swp1 ip address 10.0.0.1/30
cumulus@switch:~$ nv set interface swp1 ip address 10.0.0.2/30
cumulus@switch:~$ nv set interface swp1 ip address 2001:DB8::1/126
cumulus@switch:~$ nv config apply

In the /etc/network/interfaces file, list all IP addresses under the iface section.

auto swp1
iface swp1
    address 10.0.0.1/30
    address 10.0.0.2/30
    address 2001:DB8::1/126

The address method and address family are not mandatory; they default to inet/inet6 and static. However, you must specify inet/inet6 when you are creating DHCP or loopback interfaces.

auto lo
iface lo inet loopback

To make non-persistent changes to interfaces at runtime, use ip addr add:

cumulus@switch:~$ sudo ip addr add 10.0.0.1/30 dev swp1
cumulus@switch:~$ sudo ip addr add 2001:DB8::1/126 dev swp1

To remove an addresses from an interface, use ip addr del:

cumulus@switch:~$ sudo ip addr del 10.0.0.1/30 dev swp1
cumulus@switch:~$ sudo ip addr del 2001:DB8::1/126 dev swp1

Interface Descriptions

You can add a description (alias) to an interface.

Interface descriptions also appear in the SNMP OID IF-MIB::ifAlias

  • Interface descriptions can have a maximum of 256 characters.
  • Avoid using apostrophes or non-ASCII characters. Cumulus Linux does not parse these characters.

The following example commands create the description hypervisor_port_1 for swp1:

cumulus@switch:~$ nv set interface swp1 description hypervisor_port_1
cumulus@switch:~$ nv config apply

In the /etc/network/interfaces file, add a description using the alias keyword:

cumulus@switch:~# sudo nano /etc/network/interfaces

auto swp1
iface swp1
    alias swp1 hypervisor_port_1

Interface Commands

You can specify user commands for an interface that run at pre-up, up, post-up, pre-down, down, and post-down.

You can add any valid command in the sequence to bring an interface up or down; however, limit the scope to network-related commands associated with the particular interface. For example, it does not make sense to install a Debian package on ifup of swp1, even though it is technically possible. See man interfaces for more details.

The following examples adds a command to an interface to enable proxy ARP:

The NVUE command is not supported. 
cumulus@switch:~# sudo nano /etc/network/interfaces
auto swp1
iface swp1
    address 12.0.0.1/30
    post-up echo 1 > /proc/sys/net/ipv4/conf/swp1/proxy_arp

If your post-up command also starts, restarts, or reloads any systemd service, you must use the --no-block option with systemctl. Otherwise, that service or even the switch itself might hang after starting or restarting. For example, to restart the dhcrelay service after bringing up a VLAN, the /etc network/interfaces configuration looks like this:

auto bridge.100
iface bridge.100
    post-up systemctl --no-block restart dhcrelay.service

Source Interface File Snippets

Sourcing interface files helps organize and manage the /etc/network/interfaces file. For example:

cumulus@switch:~$ sudo cat /etc/network/interfaces
# The loopback network interface
auto lo
iface lo inet loopback

# The primary network interface
auto eth0
iface eth0 inet dhcp

source /etc/network/interfaces.d/bond0

The contents of the sourced file used above are:

cumulus@switch:~$ sudo cat /etc/network/interfaces.d/bond0
auto bond0
iface bond0
    address 14.0.0.9/30
    address 2001:ded:beef:2::1/64
    bond-slaves swp25 swp26

Port Ranges

To specify port ranges in commands:

Use commas to separate different port ranges (for example, swp1-46,10-12):

cumulus@switch:~$ nv set interface swp1-4,6,10-12 bridge domain br_default
cumulus@switch:~$ nv config apply

Use the glob keyword to specify bridge ports and bond slaves:

auto br0
iface br0
    bridge-ports glob swp1-6.100

auto br1
iface br1
    bridge-ports glob swp7-9.100  swp11.100 glob swp15-18.100

Fast Linkup

Cumulus Linux supports fast linkup on interfaces on NVIDIA Spectrum1 switches. Fast linkup enables you to bring up ports with cards that require links to come up fast, such as certain 100G optical network interface cards.

You must configure both sides of the connection with the same speed and FEC settings.

cumulus@switch:~$ nv set interface swp1 link fast-linkup on
cumulus@switch:~$ nv config apply

Edit the /etc/cumulus/switchd.conf file and add the interface.<interface>.enable_media_depended_linkup_flow=TRUE and interface.<interface>.enable_port_short_tuning=TRUE settings for the interfaces on which you want to enable fast linkup. The following example enables fast linkup on swp1:

cumulus@switch:~$ sudo nano /etc/cumulus/switchd.conf
...
interface.swp1.enable_media_depended_linkup_flow=TRUE
interface.swp1.enable_short_tuning=TRUE

Reload switchd with the sudo systemctl reload switchd.service command.

Cumulus Linux enables link flap detection by default. Link flap detection triggers when there are five link flaps within ten seconds, at which point the interface goes into a protodown state and shows linkflap as the reason. The switchd service also shows a log message similar to the following:

2023-02-10T17:53:21.264621+00:00 cumulus switchd[10109]: sync_port.c:2263 ERR swp2 link flapped more than 3 times in the last 60 seconds, setting protodown

To show interfaces with the protodown flag, run the Linux ip link command:

cumulus@switch:~$ ip link
37: swp2: <NO-CARRIER,BROADCAST,MULTICAST,SLAVE,UP> mtu 9178 qdisc pfifo_fast master bond131 state DOWN mode DEFAULT group default qlen 1000
  link/ether 1c:34:da:ba:bb:2a brd ff:ff:ff:ff:ff:ff protodown on protodown_reason <linkflap>

Clear the Interface Protodown State and Reason

The ifdown and ifup commands do not clear the protodown state. You must clear the protodown state and the reason manually using the sudo ip link set <interface> protodown_reason linkflap off and sudo ip link set <interface> protodown off commands.

cumulus@switch:~$ sudo ip link set swp2 protodown_reason linkflap off
cumulus@switch:~$ sudo ip link set swp2 protodown off

After a few seconds the port state returns to UP. Run the ip link show <interface> command to verify that the interface is no longer in a protodown state and that the reason is cleared:

cumulus@switch:~$ ip link show swp2
37: swp2: <NO-CARRIER,BROADCAST,MULTICAST,SLAVE,UP> mtu 9178 qdisc pfifo_fast master bond131 state UP mode DEFAULT group default qlen 1000
  link/ether 1c:34:da:ba:bb:2a brd ff:ff:ff:ff:ff:ff

You can change link flap protection settings in the /etc/cumulus/switchd.conf file:

  • To change the duration during which a link must flap the number of times set in the link flap threshold before link flap protection triggers, change the link_flap_window setting.
  • To change the number of times the link must flap within the link flap window before link flap protection triggers, change the link_flap_threshold setting.
  • To disable link flap protection, set the link_flap_window and link_flap_threshold parameters to 0 (zero).

After you change the link flap settings, you must restart switchd with the sudo systemctl restart switchd.service command.

Mako Templates

ifupdown2 supports Mako-style templates. The Mako template engine processes the interfaces file before parsing.

Use the template to declare cookie-cutter bridges and to declare addresses in the interfaces file:

%for i in [1,12]:
auto swp${i}
iface swp${i}
    address 10.20.${i}.3/24

  • In Mako syntax, use square brackets ([1,12]) to specify a list of individual numbers. Use range(1,12) to specify a range of interfaces.
  • To test your template and confirm it evaluates correctly, run mako-render /etc/network/interfaces.

To comment out content in Mako templates, use double hash marks (##). For example:

## % for i in range(1, 4):
## auto swp${i}
## iface swp${i}
## % endfor
##

For more Mako template examples, refer to this knowledge base article.

ifupdown Scripts

Unlike the traditional ifupdown system, ifupdown2 does not run scripts installed in /etc/network/*/ automatically to configure network interfaces.

To enable or disable ifupdown2 scripting, edit the addon_scripts_support line in the /etc/network/ifupdown2/ifupdown2.conf file. 1 enables scripting and 2 disables scripting. For example:

cumulus@switch:~$ sudo nano /etc/network/ifupdown2/ifupdown2.conf
# Support executing of ifupdown style scripts.
# Note that by default python addon modules override scripts with the same name
addon_scripts_support=1

ifupdown2 sets the following environment variables when executing commands:

  • $IFACE represents the physical name of the interface; for example, br0 or vxlan42. The name comes from the /etc/network/interfaces file.
  • $LOGICAL represents the logical name (configuration name) of the interface.
  • $METHOD represents the address method; for example, loopback, DHCP, DHCP6, manual, static, and so on.
  • $ADDRFAM represents the address families associated with the interface in a comma-separated list; for example, "inet,inet6".

Troubleshooting

To see the link and administrative state of an interface:

cumulus@switch:~$ nv show interface swp1 link state

In the following example, swp1 is administratively UP and the physical link is UP (LOWER_UP).

cumulus@switch:~$ ip link show dev swp1
3: swp1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 500
    link/ether 44:38:39:00:03:c1 brd ff:ff:ff:ff:ff:ff

To show the assigned IP address on an interface:

cumulus@switch:~$ nv show interface swp1 ip address

cumulus@switch:~$ ip addr show swp1
3: swp1: <BROADCAST,MULTICAST,SLAVE,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP qlen 500
    link/ether 44:38:39:00:03:c1 brd ff:ff:ff:ff:ff:ff
    inet 192.0.2.1/30 scope global swp1
    inet 192.0.2.2/30 scope global swp1
    inet6 2001:DB8::1/126 scope global tentative
        valid_lft forever preferred_lft forever

To show the description (alias) for an interface:

cumulus@switch$ nv show interface swp1

cumulus@switch$ ip link show swp1
3: swp1: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc pfifo_fast state DOWN mode DEFAULT qlen 500
    link/ether aa:aa:aa:aa:aa:bc brd ff:ff:ff:ff:ff:ff
    alias hypervisor_port_1

Considerations

Even though ifupdown2 supports the inclusion of multiple iface stanzas for the same interface, use a single iface stanza for each interface. If you must specify more than one iface stanza; for example, if the configuration for a single interface comes from many places, like a template or a sourced file, make sure the stanzas do not specify the same interface attributes. Otherwise, you see unexpected behavior.

In the following example, swp1 is in two files: /etc/network/interfaces and /etc/network/interfaces.d/speed_settings. ifupdown2 parses this configuration because the same attributes are not in multiple iface stanzas.

cumulus@switch:~$ sudo cat /etc/network/interfaces

source /etc/network/interfaces.d/speed_settings

auto swp1
iface swp1
  address 10.0.14.2/24

cumulus@switch:~$ cat /etc/network/interfaces.d/speed_settings

auto swp1
iface swp1
  link-speed 1000
  link-duplex full

ifupdown2 and sysctl

For sysctl commands in the pre-up, up, post-up, pre-down, down, and post-down lines that use the $IFACE variable, if the interface name contains a dot (.), ifupdown2 does not change the name to work with sysctl. For example, the interface name bridge.1 does not convert to bridge/1.

ifupdown2 and the gateway Parameter

The default route that the gateway parameter creates in ifupdown2 does not install in FRR, therefore does not redistribute into other routing protocols. Define a static default route instead, which installs in FRR and redistributes, if needed.

The following shows an example of the /etc/network/interfaces file when you use a static route instead of a gateway parameter:

auto swp2
iface swp2
address 172.16.3.3/24
up ip route add default via 172.16.3.2

Interface Name Limitations

Interface names can be a maximum of 15 characters. You cannot use a number for the first character and you cannot include a dash (-) in the name. In addition, you cannot use any name that matches with the regular expression .{0,13}\-v.*.

If you encounter issues, remove the interface name from the /etc/network/interfaces file, then restart the networking.service.

cumulus@switch:~$ sudo nano /etc/network/interfaces
cumulus@switch:~$ sudo systemctl restart networking.service

IP Address Scope

ifupdown2 does not honor the configured IP address scope setting in the /etc/network/interfaces file and treats all addresses as global. It does not report an error. Consider this example configuration:

auto swp2
iface swp2
    address 35.21.30.5/30
    address 3101:21:20::31/80
    scope link

When you run ifreload -a on this configuration, ifupdown2 considers all IP addresses as global.

cumulus@switch:~$ ip addr show swp2
5: swp2: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP group default qlen 1000
link/ether 74:e6:e2:f5:62:82 brd ff:ff:ff:ff:ff:ff
inet 35.21.30.5/30 scope global swp2
valid_lft forever preferred_lft forever
inet6 3101:21:20::31/80 scope global
valid_lft forever preferred_lft forever
inet6 fe80::76e6:e2ff:fef5:6282/64 scope link
valid_lft forever preferred_lft forever

To work around this issue, configure the IP address scope:

The NVUE command is not supported.

In the /etc/network/interfaces file, configure the IP address scope using post-up ip address add <address> dev <interface> scope <scope>. For example:

auto swp6
iface swp6
    post-up ip address add 71.21.21.20/32 dev swp6 scope site

Then run the ifreload -a command on this configuration.

The following configuration shows the correct scope:

cumulus@switch:~$ ip addr show swp6
9: swp6: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP group default qlen 1000
link/ether 74:e6:e2:f5:62:86 brd ff:ff:ff:ff:ff:ff
inet 71.21.21.20/32 scope site swp6
valid_lft forever preferred_lft forever
inet6 fe80::76e6:e2ff:fef5:6286/64 scope link
valid_lft forever preferred_lft forever

    Copyright © 2024  NVIDIA Corporation