--I heard that it has higher functionality than the mainstream bonding driver up to CentOS6. --If you want to run Docker and k8s hosts on-premises, you want to make the NIC redundant. ――Since the cloud was in full swing and I hadn't touched the on-premise server for a while, it also served as rehabilitation. -** A new on-preserver was delivered after a long time, so I wanted to play before building the production environment **
HPE DL380 Gen10 Cisco Catalyst 3750X-48 Stack Configuration
Teamd is a network teaming daemon added from CentOS 7 series instead of bonding that was major up to CentOS 6 series. First, run the following command to install Teamd:
[root@host1 ~]# dnf -y install teamd
Since CetOS7 series and later, Network Manager is used to manage the network, so the settings are also set using the command line utility Nmcli.
First, check the status of the network interface with the following command.
[root@host1 ~]# nmcli device status
DEVICE TYPE STATE CONNECTION
eno5 ethernet connected eno5
eno6 ethernet connected eno6
eno7 ethernet disconnected--
eno8 ethernet connected eno8
lo loopback No management--
Currently, the server has four physical Ethernet ports. This time I will set the teaming using eno5 and eno6.
[root@host1 ~]# nmcli connection show
NAME UUID TYPE DEVICE
eno5 c5a9f44a-306a-4728-82e8-70d0ce3e528b ethernet eno5
eno6 93be4fa2-61aa-4360-b570-94bef5f56fcb ethernet eno6
eno7 1dfffcc1-7807-4093-bdd7-844d0ce31aec ethernet --
eno8 ca899a0f-ac24-4a65-a2d6-86e90bef5f56 ethernet eno8
Delete the default settings of eno5 and eno6 by specifying the UUID with the following command.
[root@dns3 ~]# nmcli connection delete c5a9f44a-306a-4728-82e8-70d0ce3e528b
[root@dns3 ~]# nmcli connection delete 93be4fa2-61aa-4360-b570-94bef5f56fcb
Check if the connection is lost
[root@host1 ~]# nmcli connection show
NAME UUID TYPE DEVICE
eno7 1dfffcc1-7807-4093-bdd7-844d0ce31aec ethernet --
eno8 ca899a0f-ac24-4a65-a2d6-86e90bef5f56 ethernet eno8
[root@host1 ~]# nmcli device status
DEVICE TYPE STATE CONNECTION
eno7 ethernet disconnected--
eno8 ethernet connected eno8
eno5 ethernet disconnected eno5
eno6 ethernet disconnected eno6
lo loopback No management--
Create a new team interface "bond0". "name" is an option setting called runner, and there are the following types.
| runner | motion |
|---|---|
| broadcast | Data is sent on all ports |
| round-robin | Data is sent in sequence on all ports |
| active-backup | One port or link is used and the other is maintained as a backup) |
| loadbalance | Use active Tx load balancing and BPF-based Tx port selector |
| lacp | 802.Implemented 3ad link aggregation control protocol |
This time, the connection partner is a Cisco switch, so let's select LACP. When you run the command, / etc / sysconfig / network-scripts / ifcfg-team-bond0 will be created automatically. It is okay to create a file directly in the above path with an editor as before, but this time I will set it with a command to get used to nmcli.
[root@host1 ~]# nmcli connection add type team con-name bond0 ifname bond0 config '{"runner": {"name": "lacp"}}'
Check if the team interface has been created with the following command.
[root@dns3 ~]# nmcli device status
DEVICE TYPE STATE CONNECTION
bond0 team disconnected bond0
eno8 ethernet connected eno8
eno7 ethernet disconnected--
lo loopback No management--
[root@host1 ~]# nmcli connection show
NAME UUID TYPE DEVICE
bond0 2a7bb890-f677-48dc-b885-1f9ff111989d team bond0
eno8 ca899a0f-ac24-4a65-a2d6-86e90bef5f56 ethernet eno8
eno7 1dfffcc1-7807-4093-bdd7-844d0ce31aec ethernet --
[root@host1 ~]# nmcli con show bond0
connection.id: bond0
connection.uuid: 2a7bb890-f677-48dc-b885-1f9ff111989d
connection.stable-id: --
connection.type: team
connection.interface-name: bond0
connection.autoconnect:Yes
~ Omitted ~
team.config: {"runner": {"name": "lacp"}}
Recently, it seems that the terms master and slave are NG, so set the main system and sub system.
[root@host1 ~]# nmcli con add type team-slave con-name bond0-slave0 ifname eno5 master bond0
[root@host1 ~]# nmcli con add type team-slave con-name bond0-slave1 ifname eno6 master bond0
Use the following command to check if the main system and sub system can be specified.
[root@dns3 ~]# nmcli device status
DEVICE TYPE STATE CONNECTION
bond0 team connected bond0
eno8 ethernet connected eno8
eno5 ethernet connected bond0-slave0
eno6 ethernet connected bond0-slave1
eno7 ethernet disconnected--
lo loopback No management--
[root@host1 ~]# nmcli connection show
NAME UUID TYPE DEVICE
bond0 2a7bb890-f677-48dc-b885-1f9ff111989d team bond0
eno8 ca899a0f-ac24-4a65-a2d6-86e90bef5f56 ethernet eno8
bond0-slave0 c5a9f44a-306a-4728-82e8-70a8c73e528b ethernet eno5
bond0-slave1 93be4fa2-61aa-4360-b570-94fcb8cb8f94 ethernet eno6
eno7 1dfffcc1-7807-4093-bdd7-844d0ce31aec ethernet --
Set the IP address and DNS in the created team interface "bond0".
[root@host1 ~]# nmcli con mod bond0 ipv4.addresses 192.168.250.1/24
[root@host1 ~]# nmcli con mod bond0 ipv4.gateway 192.168.250.254
[root@host1 ~]# nmcli con mod bond0 ipv4.dns 192.168.250.254
[root@host1 ~]# nmcli con mod bond0 ipv4.method manual
[root@host1 ~]# nmcli con mod bond0 connection.autoconnect yes
To reflect the settings, restart the team interface "bond0" with the following command. (If it's a hassle, systemctl restart NetworkManager is probably fine too)
nmcli connection down bond0 && nmcli connection up bond0
It is normal if both link summaries are UP by executing the following command
[root@host1 ~]# teamdctl bond0 state
setup:
runner: lacp
ports:
eno5
link watches:
link summary: up
instance[link_watch_0]:
name: ethtool
link: up
down count: 0
runner:
aggregator ID: 2, Selected
selected: yes
state: current
eno6
link watches:
link summary: up
instance[link_watch_0]:
name: ethtool
link: up
down count: 0
runner:
aggregator ID: 2, Selected
selected: yes
state: current
runner:
active: yes
fast rate: no
You can check if the IP address is set with the following command.
[root@host1 ~]# ip address show dev bond0
6: bond0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
link/ether 98:f2:b3:**:**:** brd ff:ff:ff:ff:ff:ff
inet 192.168.250.1/24 brd 192.168.250.255 scope global noprefixroute bond0
valid_lft forever preferred_lft forever
inet6 fe80::**/64 scope link noprefixroute
valid_lft forever preferred_lft forever
Originally, if you describe it in Port-Channel, the setting on the interface side may be about description, but it is described just in case. LACP mode was connected with mode on in CentOS 6 bonding, but when using teamd, specify mode active. (I was addicted to this for about 5 minutes without noticing this)
interface GigabitEthernet1/0/1
description "Host1 Port#1"
switchport mode access
interface GigabitEthernet2/0/1
description "Host1 Port#2"
switchport mode access
interface Port-channel1
description "c3750_to_Host1"
switchport mode access
interface range GigabitEthernet1/0/1, GigabitEthernet2/0/1
channel-group 1 mode active
Personally, I felt it was easier than the CentOS 6 series. Redhat also recommends using tamed in the future, so I think it is better to use this.
[CentOS7] Teaming (LACP) with Catalyst 3750 and CentOS7 ｜ MSeeeeN https://mseeeen.msen.jp/teaming-lacp-with-catalyst3750-and-centos7/
How to Configure NIC Teaming on CentOS 8 / RHEL 8 Linuxtechi https://www.linuxtechi.com/configure-nic-teaming-centos-8-rhel-8/
8.3. Comparison of network teaming and bonding https://access.redhat.com/documentation/ja-jp/red_hat_enterprise_linux/7/html/networking_guide/sec-comparison_of_network_teaming_to_bonding
Recommended Posts