Linux dual-nic binding practices

Linux dual-nic binding practices
Preface

The work mainly focuses on SUSE, and the network is one of the most important links in the high-availability architecture. Physically, it is generally in the dual-nic binding mode. The default mode is usually 1 (active-backup) as the master-slave relationship.

In the production server that recently tested big data, the binding mode is set to mode = 6 (balance-alb), and the traffic is doubled by using Server Load balancer. In addition, in The SuSE HA architecture test from Skybility HA to Corosync/Openais + Pacemaker, mode = 0 (balance-rr) is used for network switch Port Channel link aggregation ), balanced round robin enables zero packet loss during network interruption.

The following uses CentOS as an example (similar to Red Hat). The dual-nic settings on other platforms are relatively simple. SuSE (YaST), AIX (SMIT), Windows (Intel ), if you have any requirements, you can reply to the message and I will update and supplement it later.

Follows the high availability principle to achieve Failure Protection and load balancing

Additional reading

• Seven Nic binding modes-

Three common Bond modes for basic configuration information

The configuration process uses mode = 6 as an example. For the other 7 modes, refer to extended reading.

Mode = 0: Load Balancing mode, with automatic backup, but requires "Switch" support and settings. Mode = 1: automatic backup mode. If one line is disconnected, other lines are automatically backed up. Mode = 6: Load Balancing mode, automatic backup, no need to Switch support and settings.

Physical Interface

Interface	IP	Type
Eth0, eth2		Two production NICs
Eth3	10.129.46.19	Private address
Bond0	10.3.3.214	Dual Nic binding address

CentOS version

1. Datanode01: ~> Cat/etc/RedHat-release
2. CentOS release 6.4 (Final)

Disable NetworkManager

1. # Disable NetworkManager immediately and disable auto-start upon startup
2. /Etc/init. d/NetworkManager stop
3. Chkconfig NetworkManager off
4. /Etc/init. d/network restart

Disable iptables and selinux (optional)

1. # Disable iptables immediately and disable auto-start upon startup
2. /Etc/init. d/iptables stop
3. Chkconfig iptables off
4.  
5. # Disable selinux immediately and disable selinux permanently
6. Setenforce 0
7. Sed-I's/SELINUX = enforcing/SELINUX = disabled/'/etc/selinux/config

Modify host name

1. Vi/etc/sysconfig/network
2.  
3. NETWORKING = yes
4. HOSTNAME = namenode01
5.  
6. # Refresh takes effect
7. Hostname namnode01
8. Source/etc/sysconfig/network

Configure the private IP Address

1. Cd/etc/sysconfig/network-scripts
2.  
3. [Root @ datanode09 network-scripts] # cat ifcfg-eth3
4. DEVICE = eth3
5. ONBOOT = yes
6. BOOTPROTO = none
7. IPADDR = 10.129.46.19
8. NETMASK = 255.255.255.0
9. IPV6INIT = no
10. USERCTL = no

Bind two NICs

1. Cd/etc/sysconfig/network-scripts
2.  
3. # Edit eth0
4. Cat> ifcfg-eth0 <EOF
5. DEVICE = eth0
6. ONBOOT = yes
7. BOOTPROTO = none
8. USERCTL = no
9. MASTER = bond0
10. EOF
11.  
12. # Edit eth2
13. Cat> ifcfg-eth2 <EOF
14. DEVICE = eth2
15. ONBOOT = yes
16. BOOTPROTO = none
17. USERCTL = no
18. MASTER = bond0
19. EOF
20.  
21. # Edit bond0
22. Cat> ifcfg-bond0 <EOF
23. DEVICE = bond0
24. TYPE = Ethernet
25. ONBOOT = yes
26. BOOTPROTO = none
27. IPADDR = 10.3.3.214
28. NETMASK = 255.255.255.0
29. GATEWAY = 10.3.3.1
30. IPV6INIT = no
31. USERCTL = no
32. EOF
33.  
34. # Set bond parameters. Pay attention to mode selection.
35. Cat>/etc/modprobe. conf <EOF
36. Alias bond0 bonding
37. Options bond0 miimon = 100 mode = 6
38. EOF
39.  
40. # Add auto-start Parameters
41. Cat>/etc/rc. local <EOF
42. Ifenslave bond0 eth0 eth2
43. EOF
44.  
45. # Restart an Eni
46. Service network restart
47.  
48. # Make the bound Nic effective immediately
49. Ifenslave bond0 eth0 eth2
50.  
51. # Test network binding
52. Ping 10.3.3.1
53.  

Comparison of three common Nic binding modes mode = 0

Interrupt any link or restore link, zero packet loss on the network

Advantage: traffic is doubled

Disadvantage: You need to connect to the same vswitch for aggregation configuration, so the high availability of the physical switch cannot be guaranteed (Cisco seems to have a solution ?)

Mode = 1

When one link is interrupted, 1-3 packets are lost (in seconds), and zero packet loss is restored.

Advantage: No need to configure the vswitch

Disadvantages:

Mode = 6

Zero packet loss occurs when any link is interrupted. 10-15 packets are lost when the link is recovered (in seconds)

Advantage: vswitches do not need to be configured, and the traffic is doubled

Disadvantage: Packet Loss takes too long to recover the link.

This article permanently updates the link address: