Troubleshooting Network Interfaces
The following sections describe various ways you can troubleshoot
Enabling Logging for Networking
/etc/default/networking file contains two settings for logging:
- To get
ifupdown2logs when the switch boots (stored in
- To enable logging when you run
systemctl [start|stop|reload] networking.service
This file also contains an option for excluding interfaces when you boot
the switch or run
systemctl start|stop|reload networking.service .
You can exclude any interface specified in
These interfaces do not come up when you boot the switch or
start/stop/reload the networking service.
cumulus@switch:~$ cat /etc/default/networking # # # Parameters for the /etc/init.d/networking script # # # Change the below to yes if you want verbose logging to be enabled VERBOSE="no" # Change the below to yes if you want debug logging to be enabled DEBUG="no" # Change the below to yes if you want logging to go to syslog SYSLOG="no" # Exclude interfaces EXCLUDE_INTERFACES=
Using ifquery to Validate and Debug Interface Configurations
ifquery to print parsed
interfaces file entries.
ifquery to pretty print
iface entries from the
cumulus@switch:~$ sudo ifquery bond0 auto bond0 iface bond0 address 22.214.171.124/30 address 2001:ded:beef:2::1/64 bond-slaves swp25 swp26
ifquery --check to check the current running state of an interface
interfaces file. It will return exit code 0 or 1 if the
configuration does not match. The line
below fails because it should read
cumulus@switch:~$ sudo ifquery --check bond0 iface bond0 bond-xmit-hash-policy layer3+7 [fail] bond-slaves swp25 swp26 [pass] address 126.96.36.199/30 [pass] address 2001:ded:beef:2::1/64 [pass]
ifquery --check is an experimental feature.
ifquery --running to print the running state of interfaces in the
interfaces file format:
cumulus@switch:~$ sudo ifquery --running bond0 auto bond0 iface bond0 bond-slaves swp25 swp26 address 188.8.131.52/30 address 2001:ded:beef:2::1/64
ifquery --syntax-help provides help on all possible attributes
supported in the
interfaces file. For complete syntax on the
interfaces file, see
man interfaces and
You can use
ifquery --print-savedstate to check the
ifdown works only on interfaces present in this state
cumulus@leaf1$ sudo ifquery --print-savedstate eth0 auto eth0 iface eth0 inet dhcp
Debugging Mako Template Errors
An easy way to debug and get details about template errors is to use the
mako-render command on your interfaces template file or on
cumulus@switch:~$ sudo mako-render /etc/network/interfaces # This file describes the network interfaces available on your system # and how to activate them. For more information, see interfaces(5). # The loopback network interface auto lo iface lo inet loopback # The primary network interface auto eth0 iface eth0 inet dhcp #auto eth1 #iface eth1 inet dhcp # Include any platform-specific interface configuration source /etc/network/interfaces.d/*.if # ssim2 added auto swp45 iface swp45 auto swp46 iface swp46 cumulus@switch:~$ sudo mako-render /etc/network/interfaces.d/<interfaces_stub_file>
ifdown Cannot Find an Interface that Exists
If you are trying to bring down an interface that you know exists, use
ifdown with the
--use-current-config option to force
check the current
/etc/network/interfaces file to find the interface.
This can solve issues where the
ifup command issues for that interface
was interrupted before it updated the state database. For example:
cumulus@switch:~$ sudo ifdown br0 error: cannot find interfaces: br0 (interface was probably never up ?) cumulus@switch:~$ sudo brctl show bridge name bridge id STP enabled interfaces br0 8000.44383900279f yes downlink peerlink cumulus@switch:~$ sudo ifdown br0 --use-current-config
Removing All References to a Child Interface
If you have a configuration with a child interface, whether it’s a VLAN, bond or another physical interface, and you remove that interface from a running configuration, you must remove every reference to it in the configuration. Otherwise, the interface continues to be used by the parent interface.
For example, consider the following configuration:
auto lo iface lo inet loopback auto eth0 iface eth0 inet dhcp auto bond1 iface bond1 bond-slaves swp2 swp1 auto bond3 iface bond3 bond-slaves swp8 swp6 swp7 auto br0 iface br0 bridge-ports swp3 swp5 bond1 swp4 bond3 bridge-pathcosts swp3=4 swp5=4 swp4=4 address 184.108.40.206/24 address 2001::10/64
Notice that bond1 is a member of br0. If bond1 is removed, you must
remove the reference to it from the br0 configuration. Otherwise, if you
reload the configuration with
ifreload -a, bond1 is still part of br0.
MTU Set on a Logical Interface Fails with Error: “Numerical result out of range”
This error occurs when the MTU you are trying to set on an interface is higher than the MTU of the lower interface or dependent interface. Linux expects the upper interface to have an MTU less than or equal to the MTU on the lower interface.
In the example below, the swp1.100 VLAN interface is an upper interface to physical interface swp1. If you want to change the MTU to 9000 on the VLAN interface, you must include the new MTU on the lower interface swp1 as well.
auto swp1.100 iface swp1.100 mtu 9000 auto swp1 iface swp1 mtu 9000
Interpreting iproute2 batch Command Failures
iproute2 commands for performance reasons. A batch
ip -force -batch - in the error message. The command
number that failed is at the end of this line:
Command failed -:1.
Below is a sample error for the command 1:
link set dev host2 master bridge. There was an error adding the bond host2 to the bridge named
bridge because host2 did not have a valid address.
error: failed to execute cmd 'ip -force -batch - [link set dev host2 master bridge addr flush dev host2 link set dev host1 master bridge addr flush dev host1 ]'(RTNETLINK answers: Invalid argument Command failed -:1) warning: bridge configuration failed (missing ports)
Understanding the “RTNETLINK answers: Invalid argument” Error when Adding a Port to a Bridge
This error can occur when the bridge port does not have a valid hardware address.
This can typically occur when the interface being added to the bridge is an incomplete bond; a bond without slaves is incomplete and does not have a valid hardware address.
MLAG Peerlink Interface Drops Many Packets
Losing a large number of packets across an MLAG peerlink interface may not be a problem. Instead this could be occurring in order to prevent looping of BUM (broadcast, unknown unicast and multicast) packets. For more information, and how to detect these drops, read the MLAG chapter.