MC-LAG (Multi-chassis link aggregation group) is multi-chassis link aggregation. It provides redundancy for connections to private network access switches and the Internet, improving infrastructure fault tolerance. For pre-configured servers, you can only set up redundancy for the private network connection. Redundancy is not available for all configurations.
You can configure MC-LAG only for servers whose configuration includes a network card with redundancy and MC-LAG.
For servers with an MC-LAG redundant connection, Servercore guarantees constant availability of one of the access switches, even during scheduled maintenance.
How it works
The server connects to two independent switches via an aggregated Ethernet channel (LAG, Etherchannel). The LACP 802.3ad protocol is used for the connection, and link aggregation is configured on the server side. In this case, two links from the access switches to the server will be active simultaneously.
10 Gbit/s — for public network, optical crossover is used for the connection;
10 Gbps — for private network, an optical cross-connect is used for connection;
25 Gbps — for private network, an optical cross-connect is used for connection.
Cost
The cost of an MC-LAG redundant connection depends on the selected connection speed.
You can check the price in the configurator on the website or when selecting server components in the Control panel.
Configure MC-LAG
Make sure your dedicated server configuration includes a network card with redundancy and that MC-LAG is added. If there is no network card with redundancy, you can order a new server with redundancy or change components for a custom server configuration.
Wait for a server readiness message from technical support. The switch ports will be aggregated into a bond.
<eth_name_1>, <eth_name_2> — names of the network interfaces to be included in the aggregation;
<ip_address> — the IP address to be used on the aggregated interface;
<mask> — subnet mask;
<gateway_4>, <gateway_6> — gateway.
Press ESC.
Exit the vi text editor, saving the changes:
:wq
Apply the new configuration:
netplan --debug apply
Verify that the bond0 network interface has bonded correctly:
cat /proc/net/bonding/bond0
Connect to the server via a network interface that will not be included in the aggregation, or via the KVM console.
Check that the bonding kernel module is installed on the server:
lsmod |grep bond
If there is no information in the response, the bonding kernel module is not installed.
If the bonding kernel module is not installed, install it:
sudo modprobe bonding
Install the package for managing and configuring parallel routing interfaces (bonding):
apt-getinstall ifenslave
Display information about network interfaces:
ifconfig-a
Disable each network interface that will be included in the aggregation one by one:
ifdown<eth_name>
Specify <eth_name> — the interface name you obtained in step 5.
Check the configuration file name of the ifupdown utility:
ls /etc/network/interfaces.d/
The response will show the configuration file name. For example:
50-cloud-init
Copy the configuration file name.
Open the configuration file of the ifupdown utility in the text editor vi:
vi /etc/network/interfaces.d/<file_name>
Specify <file_name> — the configuration file name you copied in step 7.
Set the settings for the network interfaces that will be included in the aggregation to the following:
auto lo
iface lo inet loopback
auto <eth_name_1>
iface <eth_name_1> inet static
bond-master bond0
bond-primary <eth_name_1> <eth_name_2>
auto <eth_name_2>
iface <eth_name_2> inet manual
bond-master bond0
bond-primary <eth_name_1> <eth_name_2>
auto bond0
iface bond0 inet static
bond-slaves <eth_name_1> <eth_name_2>
bond-mode 802.3ad
bond-miimon 100
bond-downdelay 100
bond-updelay 100
bond-xmit-hash-policy layer3+4
address <ip_address>
netmask <mask>
gateway <gateway>
dns-nameservers <dns_servers>
Specify:
<eth_name_1>, <eth_name_2> — names of the network interfaces to be included in the aggregation;
<ip_address> — the IP address to be used on the aggregated interface;
<mask> — subnet mask;
<gateway> — gateway;
<dns_servers> — DNS server address. We recommend using Servercore recursive DNS servers, but you can specify any available DNS servers.
Press ESC.
Exit the vi text editor, saving the changes:
:wq
Apply the network configuration changes:
source /etc/network/interfaces.d/<file_name>
Specify <file_name> — the configuration file name you obtained in step 7.
Bring up the bond0 network interface:
ifup bond0
Restart the network services:
/etc/init.d/networking start
Verify that the bond0 network interface has bonded correctly:
cat /proc/net/bonding/bond0
In Windows Server 2019, you can combine multiple network interfaces into a single logical interface using NIC Teaming.
Server Manager
PowerShell
Connect to the server via a network interface that will not be included in the aggregation, or via the KVM console.
Launch Server Manager.
Go to Local Server → Properties.
Click NIC Teaming.
In the Servers section, select the server to configure.
In the Groups section, click Tasks and select New Team.
In the Team name field, enter the team name.
In the Member adapters section, select the network adapters to add to the team.
In the Teaming mode field, select LACP.
In the Load balancing mode field, select a load balancing algorithm.
Optional: in the Primary team interface field, enter the VLAN ID for the team interface if it is used in a private network and you have Q-in-Q enabled. Do not use a VLAN ID for a public network interface.
Starting with Windows Server 2022, NIC Teaming is replaced by Switch Embedded Teaming (SET). SET can only be configured when creating a Hyper-V virtual switch.
Connect to the server via a network interface that will not be included in the aggregation, or via the KVM console.