Bond in active-backup mode does not switch to backup interface when the primary link goes down

[cotosso]

Version: NethSecurity 8.7.1

Brief description
When using a bond interface in active backup mode, the system does not switch to the secondary interface when the primary interface becomes unavailable. The failure occurs because no link monitoring is configured for the bond.

Steps to reproduce

• Configure a bond in active-backup mode
• send traffic over the bond (e.g. a ping)
• Disconnect the primary slave

Expected behavior

• the traffic will start to flow in the backup slave

Actual behavior

• the traffic stops to flow

Why this is a problem
Active backup mode requires link monitoring to detect when the active interface goes down. Without it, the system continues to consider the primary interface available even when it has failed. This leads to loss of connectivity and makes the bond unreliable.

Proposed solution
Add link monitoring options to the bond configuration. The required options are:

option link_monitoring 'mii'
option miimon '100'

These options enable MII monitoring every 100 milliseconds and allow the bond to detect link-down events and switch to the backup interface.
Based on typical Linux bonding behavior, these options are standard and usually harmless even when not strictly required, but their safety in every NethSecurity context should be confirmed during implementation.

Components

NethSecurity 8.7.1

[m-dilorenzi]

QA

UCI default script

The new script 99-nethsec-bond-link-monitoring should update an existing bond to set the default link-monitoring mode.

• Create a bond called test_bond in active-backup mode using the 8.7.1 stable image.
• Check the bond configuration on the firewall. It should look like the following (except for the interfaces used and the bond IP):

network.test_bond=interface
network.test_bond.proto='bonding'
network.test_bond.ipaddr='127.35.37.1'
network.test_bond.netmask='255.255.255.255'
network.test_bond.slaves='eth2' 'eth3'
network.test_bond.bonding_policy='active-backup'
network.test_bond.primary='eth2'
network.test_bond.primary_reselect='always'
network.test_bond.fail_over_mac='none'
network.test_bond.num_grat_arp__num_unsol_na='1'

• Download the new image.
• Update the firewall with the downloaded image.
• Check the bond configuration; it should be updated with the default link-monitoring properties:

network.test_bond.miimon='100'
network.test_bond.link_monitoring='mii'

UCI default script with existing link monitoring options

The new script 99-nethsec-bond-link-monitoring should update an existing bond to set the default link-monitoring mode only if no link monitoring properties have been already defined.

• Create a bond called test_bond in active-backup mode using the 8.7.1 stable image.
• Check the bond configuration on the firewall. It should look like the following (except for the interfaces used and the bond IP):

network.test_bond=interface
network.test_bond.proto='bonding'
network.test_bond.ipaddr='127.35.37.1'
network.test_bond.netmask='255.255.255.255'
network.test_bond.slaves='eth2' 'eth3'
network.test_bond.bonding_policy='active-backup'
network.test_bond.primary='eth2'
network.test_bond.primary_reselect='always'
network.test_bond.fail_over_mac='none'
network.test_bond.num_grat_arp__num_unsol_na='1'

• add the link monitoring options using the following commands:

uci set network.test_bond.link_monitoring='mii'
uci set network.test_bond.miimon='200'
uci commit

You can also specify different link monitoring options, for example arp_interval:

uci set network.test_bond.arp_interval='100'
uci set network.test_bond.arp_ip_target='{recheable_host_inside_network}'
uci commit

• Download the new image.
• Update the firewall with the downloaded image.
• Check the bond configuration; it should remain the same with no updates with the default link-monitoring properties:

network.test_bond.miimon='200'
network.test_bond.link_monitoring='mii'

or with arp_interval as well.

New bond creation

When a new bond is created using the latest image, it should be created with the link-monitoring properties.

• Create a bond called test_bond.
• Check the bond configuration on the firewall. It should look like the following (except for the interfaces used and the bond IP):

network.test_bond=interface
network.test_bond.proto='bonding'
network.test_bond.ipaddr='127.29.220.1'
network.test_bond.netmask='255.255.255.255'
network.test_bond.link_monitoring='mii'
network.test_bond.miimon='100'
network.test_bond.slaves='eth2' 'eth3'
network.test_bond.bonding_policy='active-backup'
network.test_bond.primary='eth2'
network.test_bond.primary_reselect='always'
network.test_bond.fail_over_mac='none'
network.test_bond.num_grat_arp__num_unsol_na='1'

with the two properties network.test_bond.link_monitoring and network.test_bond.miimon already set.

• Check the active interface on the bond using:

cat /proc/net/bonding/bond-test_bond

and verify that the active interface is the primary you set (in the example, eth2):

...
Currently Active Slave: eth2
...

• Simulate the primary interface going down with:

ip link set dev eth2 down

• Check the active interface again with:

cat /proc/net/bonding/bond-test_bond

You should see:

...
Currently Active Slave: eth3
...

[cotosso]

All test cases went fine and the results were exactly the same expected from test cases description.

Test case 1: UCI default script : OK

Test case 2: UCI default script with existing link monitoring options : OK

Test case 3 : New bond creation : OK