From Fedora Project Wiki
Line 22: Line 22:
The purpose of the Team driver is to provide a mechanism to team multiple NICs (ports) into one logical one (teamdev) at L2 layer. The process is called "channel bonding", "Ethernet bonding", "channel teaming", "link aggregation", etc. This is already implemented in the Linux kernel by the bonding driver. The main thing to realize is that the Team driver project is not trying to replicate or mimic the bonding driver. What it does is it resolve the same problem using a different approach. Therefore, for example, the way Team is configured differs dramatically from the way bonding is. Team has many advantages over Bonding. It's modular, userspace driven, fast.
The purpose of the Team driver is to provide a mechanism to team multiple NICs (ports) into one logical one (teamdev) at L2 layer. The process is called "channel bonding", "Ethernet bonding", "channel teaming", "link aggregation", etc. This is already implemented in the Linux kernel by the bonding driver. The main thing to realize is that the Team driver project is not trying to replicate or mimic the bonding driver. What it does is it resolve the same problem using a different approach. Therefore, for example, the way Team is configured differs dramatically from the way bonding is. Team has many advantages over Bonding. It's modular, userspace driven, fast.


This feature consists of two parts: kernel module (modules) and userspace library+daemon (libteam package).
This feature consists of two parts:
* kernel module (modules). They are already part of Fedora 18 kernel package.
* userspace library (libteam package) and daemon (teamd package).


== Benefit to Fedora ==
== Benefit to Fedora ==

Revision as of 11:32, 19 July 2012

= Network Team driver

Summary

Network Team driver allows multiple NICs to be teamed together and act like a single one.

Owner

Current status

  • Targeted release: Fedora 18
  • Last updated: 2012-07-18
  • Percentage of completion: 90%


Detailed Description

The purpose of the Team driver is to provide a mechanism to team multiple NICs (ports) into one logical one (teamdev) at L2 layer. The process is called "channel bonding", "Ethernet bonding", "channel teaming", "link aggregation", etc. This is already implemented in the Linux kernel by the bonding driver. The main thing to realize is that the Team driver project is not trying to replicate or mimic the bonding driver. What it does is it resolve the same problem using a different approach. Therefore, for example, the way Team is configured differs dramatically from the way bonding is. Team has many advantages over Bonding. It's modular, userspace driven, fast.

This feature consists of two parts:

  • kernel module (modules). They are already part of Fedora 18 kernel package.
  • userspace library (libteam package) and daemon (teamd package).

Benefit to Fedora

This will bring possibility to run userspace driven NIC teaming. There are some features not supported by bonding are supported by Team driver, for example:

  • load balancing for LACP
  • separate per-port link monitoring setup
  • port priorities and stickiness
  • IPv6 Neighbor Solicitation/Neighbor Advertisement link monitoring

Scope

This is very isolated. It should not affect anything else. It does not touch any existing things.

How To Test

Install "teamd" package first. You need two network interfaces for this basic test. In case their names are not "eth1" and "eth2", please adjust the config file.

test2:~# ip link
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 16436 qdisc noqueue state UNKNOWN mode DEFAULT 
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 1000
    link/ether 52:54:00:1a:db:bb brd ff:ff:ff:ff:ff:ff
3: eth1: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN mode DEFAULT qlen 1000
    link/ether 52:54:00:99:bb:27 brd ff:ff:ff:ff:ff:ff
4: eth2: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN mode DEFAULT qlen 1000
    link/ether 52:54:00:b7:cc:fb brd ff:ff:ff:ff:ff:ff
test2:~# cat /usr/share/doc/
Display all 194 possibilities? (y or n)
test2:~# cat /usr/share/doc/teamd-0.1/example_configs/activebackup_ethtool_1.conf 
{
        "device":       "team0",
        "runner":       {"name": "activebackup"},
        "link_watch":   {"name": "ethtool"},
        "ports":        {
                "eth1": {
                        "prio": -10,
                        "sticky": true
                },
                "eth2": {
                        "prio": 100
                }
        }
}
test2:~# teamd -f /usr/share/doc/teamd-0.1/example_configs/activebackup_ethtool_1.conf -d
test2:~# ip link
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 16436 qdisc noqueue state UNKNOWN mode DEFAULT 
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 1000
    link/ether 52:54:00:1a:db:bb brd ff:ff:ff:ff:ff:ff
3: eth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast master team0 state UP mode DEFAULT qlen 1000
    link/ether ea:8e:85:d3:95:5d brd ff:ff:ff:ff:ff:ff
4: eth2: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast master team0 state UP mode DEFAULT qlen 1000
    link/ether 52:54:00:b7:cc:fb brd ff:ff:ff:ff:ff:ff
5: team0: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN mode DEFAULT 
    link/ether ea:8e:85:d3:95:5d brd ff:ff:ff:ff:ff:ff
test2:~# ip addr add 192.168.23.2/24 dev team0
test2:~# ip link set team0 up
test2:~# ip link
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 16436 qdisc noqueue state UNKNOWN mode DEFAULT 
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 1000
    link/ether 52:54:00:1a:db:bb brd ff:ff:ff:ff:ff:ff
3: eth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast master team0 state UP mode DEFAULT qlen 1000
    link/ether ea:8e:85:d3:95:5d brd ff:ff:ff:ff:ff:ff
4: eth2: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast master team0 state UP mode DEFAULT qlen 1000
    link/ether 52:54:00:b7:cc:fb brd ff:ff:ff:ff:ff:ff
5: team0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP mode DEFAULT 
    link/ether ea:8e:85:d3:95:5d brd ff:ff:ff:ff:ff:ff
test2:~# teamd -f /usr/share/doc/teamd-0.1/example_configs/activebackup_ethtool_1.conf -k
test2:~# ip link
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 16436 qdisc noqueue state UNKNOWN mode DEFAULT 
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 1000
    link/ether 52:54:00:1a:db:bb brd ff:ff:ff:ff:ff:ff
3: eth1: <BROADCAST,MULTICAST> mtu 1500 qdisc pfifo_fast state DOWN mode DEFAULT qlen 1000
    link/ether 52:54:00:99:bb:27 brd ff:ff:ff:ff:ff:ff
4: eth2: <BROADCAST,MULTICAST> mtu 1500 qdisc pfifo_fast state DOWN mode DEFAULT qlen 1000
    link/ether 52:54:00:b7:cc:fb brd ff:ff:ff:ff:ff:ff

User Experience

User will not experience any difference comparing to existing state. Only there is new possibility to create Team devices.

Dependencies

No dependencies.

Contingency Plan

Currently, the feature is fully functional. Only some little things and tweaks are going to go in for Fedora 18. But if some of this little things won't make it, it would not be a show-stopper.

Documentation

Release Notes

Fedora 18 includes Team Driver which provides a possibility to team multiple network interfaces together so they act like a single network interface. This extends teaming possibilities provided by existing bonding driver. It also provides userspace driven, modular alternative to bonding driver.

Comments and Discussion