Merge Bonding form Wiki
[pve-docs.git] / pve-network.adoc
1 Network Configuration
2 ---------------------
3 include::attributes.txt[]
4
5 {pve} uses a bridged networking model. Each host can have up to 4094
6 bridges. Bridges are like physical network switches implemented in
7 software. All VMs can share a single bridge, as if
8 virtual network cables from each guest were all plugged into the same
9 switch. But you can also create multiple bridges to separate network
10 domains.
11
12 For connecting VMs to the outside world, bridges are attached to
13 physical network cards. For further flexibility, you can configure
14 VLANs (IEEE 802.1q) and network bonding, also known as "link
15 aggregation". That way it is possible to build complex and flexible
16 virtual networks.
17
18 Debian traditionally uses the `ifup` and `ifdown` commands to
19 configure the network. The file `/etc/network/interfaces` contains the
20 whole network setup. Please refer to to manual page (`man interfaces`)
21 for a complete format description.
22
23 NOTE: {pve} does not write changes directly to
24 `/etc/network/interfaces`. Instead, we write into a temporary file
25 called `/etc/network/interfaces.new`, and commit those changes when
26 you reboot the node.
27
28 It is worth mentioning that you can directly edit the configuration
29 file. All {pve} tools tries hard to keep such direct user
30 modifications. Using the GUI is still preferable, because it
31 protect you from errors.
32
33
34 Naming Conventions
35 ~~~~~~~~~~~~~~~~~~
36
37 We currently use the following naming conventions for device names:
38
39 * Ethernet devices: eth[N], where 0 ≤ N (`eth0`, `eth1`, ...)
40
41 * Bridge names: vmbr[N], where 0 ≤ N ≤ 4094 (`vmbr0` - `vmbr4094`)
42
43 * Bonds: bond[N], where 0 ≤ N (`bond0`, `bond1`, ...)
44
45 * VLANs: Simply add the VLAN number to the device name,
46 separated by a period (`eth0.50`, `bond1.30`)
47
48 This makes it easier to debug networks problems, because the device
49 names implies the device type.
50
51 Default Configuration using a Bridge
52 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
53
54 The installation program creates a single bridge named `vmbr0`, which
55 is connected to the first ethernet card `eth0`. The corresponding
56 configuration in `/etc/network/interfaces` looks like this:
57
58 ----
59 auto lo
60 iface lo inet loopback
61
62 iface eth0 inet manual
63
64 auto vmbr0
65 iface vmbr0 inet static
66 address 192.168.10.2
67 netmask 255.255.255.0
68 gateway 192.168.10.1
69 bridge_ports eth0
70 bridge_stp off
71 bridge_fd 0
72 ----
73
74 Virtual machines behave as if they were directly connected to the
75 physical network. The network, in turn, sees each virtual machine as
76 having its own MAC, even though there is only one network cable
77 connecting all of these VMs to the network.
78
79
80 Routed Configuration
81 ~~~~~~~~~~~~~~~~~~~~
82
83 Most hosting providers do not support the above setup. For security
84 reasons, they disable networking as soon as they detect multiple MAC
85 addresses on a single interface.
86
87 TIP: Some providers allows you to register additional MACs on there
88 management interface. This avoids the problem, but is clumsy to
89 configure because you need to register a MAC for each of your VMs.
90
91 You can avoid the problem by ``routing'' all traffic via a single
92 interface. This makes sure that all network packets use the same MAC
93 address.
94
95 A common scenario is that you have a public IP (assume `192.168.10.2`
96 for this example), and an additional IP block for your VMs
97 (`10.10.10.1/255.255.255.0`). We recommend the following setup for such
98 situations:
99
100 ----
101 auto lo
102 iface lo inet loopback
103
104 auto eth0
105 iface eth0 inet static
106 address 192.168.10.2
107 netmask 255.255.255.0
108 gateway 192.168.10.1
109 post-up echo 1 > /proc/sys/net/ipv4/conf/eth0/proxy_arp
110
111
112 auto vmbr0
113 iface vmbr0 inet static
114 address 10.10.10.1
115 netmask 255.255.255.0
116 bridge_ports none
117 bridge_stp off
118 bridge_fd 0
119 ----
120
121
122 Masquerading (NAT) with `iptables`
123 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
124
125 In some cases you may want to use private IPs behind your Proxmox
126 host's true IP, and masquerade the traffic using NAT:
127
128 ----
129 auto lo
130 iface lo inet loopback
131
132 auto eth0
133 #real IP adress
134 iface eth0 inet static
135 address 192.168.10.2
136 netmask 255.255.255.0
137 gateway 192.168.10.1
138
139 auto vmbr0
140 #private sub network
141 iface vmbr0 inet static
142 address 10.10.10.1
143 netmask 255.255.255.0
144 bridge_ports none
145 bridge_stp off
146 bridge_fd 0
147
148 post-up echo 1 > /proc/sys/net/ipv4/ip_forward
149 post-up iptables -t nat -A POSTROUTING -s '10.10.10.0/24' -o eth0 -j MASQUERADE
150 post-down iptables -t nat -D POSTROUTING -s '10.10.10.0/24' -o eth0 -j MASQUERADE
151 ----
152
153
154 Linux Bond
155 ~~~~~~~~~~
156
157 Bonding (also called NIC teaming or Link Aggregation) is a technique
158 for binding multiple NIC's to a single network device. It is possible
159 to achieve different goals, like make the network fault-tolerant,
160 increase the performance or both together.
161
162 High-speed hardware like Fibre Channel and the associated switching
163 hardware can be quite expensive. By doing link aggregation, two NICs
164 can appear as one logical interface, resulting in double speed. This
165 is a native Linux kernel feature that is supported by most
166 switches. If your nodes have multiple Ethernet ports, you can
167 distribute your points of failure by running network cables to
168 different switches and the bonded connection will failover to one
169 cable or the other in case of network trouble.
170
171 Aggregated links can improve live-migration delays and improve the
172 speed of replication of data between Proxmox VE Cluster nodes.
173
174 There are 7 modes for bonding:
175
176 * *Round-robin (balance-rr):* Transmit network packets in sequential
177 order from the first available network interface (NIC) slave through
178 the last. This mode provides load balancing and fault tolerance.
179
180 * *Active-backup (active-backup):* Only one NIC slave in the bond is
181 active. A different slave becomes active if, and only if, the active
182 slave fails. The single logical bonded interface's MAC address is
183 externally visible on only one NIC (port) to avoid distortion in the
184 network switch. This mode provides fault tolerance.
185
186 * *XOR (balance-xor):* Transmit network packets based on [(source MAC
187 address XOR'd with destination MAC address) modulo NIC slave
188 count]. This selects the same NIC slave for each destination MAC
189 address. This mode provides load balancing and fault tolerance.
190
191 * *Broadcast (broadcast):* Transmit network packets on all slave
192 network interfaces. This mode provides fault tolerance.
193
194 * *IEEE 802.3ad Dynamic link aggregation (802.3ad)(LACP):* Creates
195 aggregation groups that share the same speed and duplex
196 settings. Utilizes all slave network interfaces in the active
197 aggregator group according to the 802.3ad specification.
198
199 * *Adaptive transmit load balancing (balance-tlb):* Linux bonding
200 driver mode that does not require any special network-switch
201 support. The outgoing network packet traffic is distributed according
202 to the current load (computed relative to the speed) on each network
203 interface slave. Incoming traffic is received by one currently
204 designated slave network interface. If this receiving slave fails,
205 another slave takes over the MAC address of the failed receiving
206 slave.
207
208 * *Adaptive load balancing (balanceIEEE 802.3ad Dynamic link
209 aggregation (802.3ad)(LACP):-alb):* Includes balance-tlb plus receive
210 load balancing (rlb) for IPV4 traffic, and does not require any
211 special network switch support. The receive load balancing is achieved
212 by ARP negotiation. The bonding driver intercepts the ARP Replies sent
213 by the local system on their way out and overwrites the source
214 hardware address with the unique hardware address of one of the NIC
215 slaves in the single logical bonded interface such that different
216 network-peers use different MAC addresses for their network packet
217 traffic.
218
219 For the most setups the active-backup are the best choice or if your
220 switch support LACP "IEEE 802.3ad" this mode should be preferred.
221
222 The following bond configuration can be used as distributed/shared
223 storage network. The benefit would be that you get more speed and the
224 network will be fault-tolerant.
225
226 .Example: Use bond with fixed IP address
227 ----
228 auto lo
229 iface lo inet loopback
230
231 iface eth1 inet manual
232
233 iface eth2 inet manual
234
235 auto bond0
236 iface bond0 inet static
237 slaves eth1 eth2
238 address 192.168.1.2
239 netmask 255.255.255.0
240 bond_miimon 100
241 bond_mode 802.3ad
242 bond_xmit_hash_policy layer2+3
243
244 auto vmbr0
245 iface vmbr0 inet static
246 address 10.10.10.2
247 netmask 255.255.255.0
248 gateway 10.10.10.1
249 bridge_ports eth0
250 bridge_stp off
251 bridge_fd 0
252
253 ----
254
255
256 Another possibility it to use the bond directly as bridge port.
257 This can be used to make the guest network fault-tolerant.
258
259 .Example: Use a bond as bridge port
260 ----
261 auto lo
262 iface lo inet loopback
263
264 iface eth1 inet manual
265
266 iface eth2 inet manual
267
268 auto bond0
269 iface bond0 inet maunal
270 slaves eth1 eth2
271 bond_miimon 100
272 bond_mode 802.3ad
273 bond_xmit_hash_policy layer2+3
274
275 auto vmbr0
276 iface vmbr0 inet static
277 address 10.10.10.2
278 netmask 255.255.255.0
279 gateway 10.10.10.1
280 bridge_ports bond0
281 bridge_stp off
282 bridge_fd 0
283
284 ----
285
286 ////
287 TODO: explain IPv6 support?
288 TODO: explan OVS
289 ////