Skip to content

fix: remove irrelevant active-backup bonding #207

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Frostman merged 1 commit into from
Oct 31, 2025
Merged

fix: remove irrelevant active-backup bonding #207

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
121 changes: 0 additions & 121 deletions docs/user-guide/host-settings.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -57,127 +57,6 @@ To check the status of this bond:
cat /proc/net/bonding/bond0
```


## LAG, Bonding, Teaming, Bundling

A LAG configuration can be used to increase the bandwidth to and from an end
host. For example, if a server has two 25 Gbps ports, the ports can be bonded so
they will have an aggregate 50 Gbps of available bandwidth. At the physical
layer, two cables are going from the server to a single switch.

To enable fast detection of faults, `802.3ad` is the supported mode for bonds
attaching to the Open Network Fabric. The steps to configure the bond are the
same as the [above server settings](#server-settings)

The kernel bonding driver documentation discusses other modes for bonding.
Several of those modes don't require cooperation on the switch side, so there
is no fabric configuration needed. The primary concerns for using these modes
are out-of-order TCP delivery, fault detection, and correction time.

## Active / Passive Fault Tolerance

Some networking equipment doesn't support MCLAG or ESLAG. In this case, to
enable fault-tolerant availability, configure a bond on the Linux server in
active / standby mode. In this topology, a link is connected from
server-1 to switch-1, and another link is connected from server-1 to switch-2.
There is neither [MCLAG connection](connections.md#mclag) nor [switch
group](devices.md#redundancy-groups) configuration applied to the
switches. When adding these links to the fabric via the `kubectl` command or
the wiring diagram, use the [unbundled](connections.md#unbundled) connection
kind.

### Active / Passive Bond Settings

The IP addresses used in the following commands are examples. On the host:

1. Create a bond:

``` bash
sudo ip link add bond0 type bond mode active-backup miimon 50 num_grat_arp 5 primary enp2s1
```
2. Add the member interfaces to the bond:

``` bash
sudo ip link set link enp2s1 master bond0
sudo ip link set link enp2s2 master bond0
```
3. Add an address to the bond:

``` bash
sudo ip addr add 10.30.10.10/24 dev bond0
```
4. Make the bond the default route:

```bash
sudo ip route add default via 10.30.10.1
```

Print the link layer information for the interfaces to see that the same MAC
address is on all 3 interfaces: bond0, enp2s1,enp2s2:

``` console
core@server-1 ~ $ ip link
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
2: enp2s1: <BROADCAST,MULTICAST,SLAVE,UP,LOWER_UP> mtu 1500 qdisc mq master bond0 state UP mode DEFAULT group default qlen 1000
link/ether a6:eb:64:de:95:92 brd ff:ff:ff:ff:ff:ff permaddr a0:36:9f:3f:1a:e8
3: enp2s2: <BROADCAST,MULTICAST,SLAVE,UP,LOWER_UP> mtu 1500 qdisc mq master bond0 state UP mode DEFAULT group default qlen 1000
link/ether a6:eb:64:de:95:92 brd ff:ff:ff:ff:ff:ff permaddr 14:02:ec:7f:3a:3c
4: enp2s0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP mode DEFAULT group default qlen 1000
link/ether 0c:20:12:fe:01:00 brd ff:ff:ff:ff:ff:ff
10: bond0: <BROADCAST,MULTICAST,MASTER,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP mode DEFAULT group default qlen 1000
link/ether a6:eb:64:de:95:92 brd ff:ff:ff:ff:ff:ff
13: bond0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP mode DEFAULT group default qlen 1000
link/ether a6:eb:64:de:95:92 brd ff:ff:ff:ff:ff:ff
```

Confirm that the bonding driver is reporting the selected primary interface:

```console
core@server-1 ~ $ cat /proc/net/bonding/bond0

Ethernet Channel Bonding Driver: v6.6.74-flatcar

Bonding Mode: fault-tolerance (active-backup)
Primary Slave: enp2s1 (primary_reselect always)
Currently Active Slave: enp2s1
MII Status: up
MII Polling Interval (ms): 50
Up Delay (ms): 0
Down Delay (ms): 0
Peer Notification Delay (ms): 0

Slave Interface: enp2s1
MII Status: up
Speed: 10000 Mbps
Duplex: full
Link Failure Count: 0
Permanent HW addr: a0:36:9f:3f:1a:e8
Slave queue ID: 0

Slave Interface: enp2s2
MII Status: up
Speed: 10000 Mbps
Duplex: full
Link Failure Count: 0
Permanent HW addr: 14:02:ec:7f:3a:3c
Slave queue ID: 0
```

### Failure Scenarios
When `miimon` detects that the primary member interface is no longer `up`, the bonding
driver will send five gratuitous ARP messages via the secondary member interface to notify the switch
that the IP address is now on a new switch. The fabric will start routing
traffic to this location, and the host will start receiving traffic.

If the `miimon` polling determines that the primary member interface link is
`up`, then the bonding driver will switch back to the primary member interface.
The `primary_reselect` behavior is configurable, see the [kernel bonding driver][bonding] documentation
for additional details.

The changeover process will result in dropped packets, but it should not result
in TCP timeouts.

## VLANs

A Linux host is capable of receiving tagged and untagged traffic. Tagged
Expand Down