Skip to content

Commit d8884ac

Browse files
yaocw2020vickyhella
yaocw2020
and
vickyhella
committed
Add network best practice document
Signed-off-by: yaocw2020 <[email protected]> Co-authored-by: vickyhella <[email protected]>
1 parent 0496dae commit d8884ac

File tree

3 files changed

+217
-0
lines changed

3 files changed

+217
-0
lines changed

‎docs/networking/best-practice.md

Lines changed: 200 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,200 @@
1+
---
2+
sidebar_position: 6
3+
sidebar_label: Best Practice
4+
title: "Harvester Network Best Practice"
5+
keywords:
6+
- Harvester
7+
- Networking
8+
---
9+
10+
<head>
11+
<link rel="canonical" href="https://docs.harvesterhci.io/v1.2/networking/best-pratice"/>
12+
</head>
13+
14+
## Overview
15+
This best practice guide introduces how to configure Harvester and the external network to achieve the following goals:
16+
- Traffic isolation between the management plane and the data plane
17+
- General external switch and router configurations
18+
- Network access to VMs from different VLANs
19+
- Access Harvester load balancers from different VLANs
20+
21+
We will use the following diagram as an example to illustrate the best practice.
22+
23+
![](/img/v1.2/networking/best-practice.png)
24+
25+
The diagram shows a Harvester cluster composed of two hosts. It contains:
26+
- Hardware:
27+
- Two Harvester servers with dual-port network cards.
28+
- One non-VLAN-aware switch and one VLAN-aware switch. We will use the Cisco-like configuration as an example.
29+
- One router. We will use the Cisco-like configuration as an example.
30+
31+
- Cabling:
32+
- The NIC eth0 of the node1 is connected to the port `ethernet1/1` of the switch1, while the NIC eth0 of the node2 is connected to the port `ethernet1/2` of the switch1.
33+
- The NIC eth1 of the node1 is connected to the port `ethernet1/1` of the switch2, while the NIC eth1 of the node2 is connected to the port `ethernet1/2` of the switch2.
34+
- The port `ethernet1/3` of the switch1 is connected to the port `ethernet0/1` of the router.
35+
- The port `ethernet1/3` of the switch2 is connected to the port `ethernet0/2` of the router.
36+
37+
- Network specification:
38+
- The subnet of the Harvester hosts is in the VLAN untagged network.
39+
- All hosts are in the IPv4 subnet `10.10.0.0/24`, and the gateway IP address is `10.10.0.254`.
40+
- The VM network allows VLAN 100-200.
41+
- The IPv4 subnets of the VM network are:
42+
- untagged network: `192.168.0.0/24`, and the gateway IP address is `192.168.0.254`.
43+
- VLAN 100: `192.168.100.0/24`, and the gateway IP address is `192.168.100.254`.
44+
- VLAN 200: `192.168.200.0/24`, and the gateway IP address is `192.168.200.254`.
45+
46+
- Harvester configuration:
47+
- Two cluster networks: `mgmt` and `vm`.
48+
- Three VM networks under the cluster network `vm`: `vlan100`, `vlan200`, and `untagged`.
49+
- Six VMs, from `VM1` to `VM6`.
50+
- One guest cluster `demo` composed of `VM3` and `VM4`.
51+
- Two VM load balancers and one guest Kubernetes cluster load balancer.
52+
53+
## Multiple Cluster Networks for Traffic Isolation
54+
The two Harvester hosts are equipped with two NICs. Specifically, NIC `eth0` is used for the management network (mapped to the cluster network `mgmt`), while NIC `eth1` is used for the VM network (mapped to the cluster network `vm`).
55+
56+
It's beneficial to use two cluster networks to achieve traffic isolation between the management plane and the data plane. If there is an issue with the VM network, you can still use the management network for emergency handling to ensure business continuity. Similarly, if there is a failure in the management network, VM traffic is not affected.
57+
58+
If your hardware is equipped with more NICs, it's recommended that you use at least two NICs for one cluster network. For example, you can use NIC `eth0` and `eth1` for the management network, and use NIC `eth2` and `eth3` for the VM network.
59+
60+
## External Switch and Router Configuration
61+
1. ** Switch1 configuration**:
62+
63+
Since the management network is under the untagged network, switch1 can be a non-VLAN-aware switch. Typically, a non-VLAN-aware switch cannot be configured.
64+
65+
2. ** Switch2 configuration**:
66+
67+
Set the ports `ethernet1/1`, `ethernet1/2`, and `ethernet1/3` as trunk ports, and allow VLAN 100-200.
68+
69+
```
70+
switch2# config terminal
71+
switch2(config)# interface ethernet1/1
72+
switch2(config-if)# switchport
73+
switch2(config-if)# switchport mode trunk
74+
switch2(config-if)# switchport trunk allowed vlan 100-200
75+
switch2(config-if)# switchport trunk native vlan 1
76+
switch2(config-if)# no shutdown
77+
switch2(config)# interface ethernet1/2
78+
switch2(config-if)# switchport
79+
switch2(config-if)# switchport mode trunk
80+
switch2(config-if)# switchport trunk allowed vlan 100-200
81+
switch2(config-if)# switchport trunk native vlan 1
82+
switch2(config-if)# no shutdown
83+
switch2(config)# interface ethernet1/3
84+
switch2(config-if)# switchport
85+
switch2(config-if)# switchport mode trunk
86+
switch2(config-if)# switchport trunk allowed vlan 100-200
87+
switch2(config-if)# switchport trunk native vlan 1
88+
switch2(config-if)# no shutdown
89+
switch2(config-if)# end
90+
switch2# copy running-config startup-config
91+
```
92+
93+
3. **Router configuration**:
94+
95+
- Configure a DHCP pool for the management network.
96+
97+
```
98+
router# config terminal
99+
router(config)# ip dhcp pool mgmt
100+
router(dhcp-config)# network 10.10.0.0 255.255.255.0
101+
router(dhcp-config)# default-router 10.10.0.254
102+
router(dhcp-config)# interface ethernet0/1
103+
router(config-if)# ip address 10.10.0.254 255.255.255.0
104+
router(config-if)# no shutdown
105+
router(config)# exit
106+
router# copy running-config startup-config
107+
```
108+
109+
- Configure three DHCP pools for the VM networks (untagged, vlan100, and vlan200).
110+
111+
```
112+
router# config terminal
113+
router(config)# ip dhcp pool vm-untagged
114+
router(dhcp-config)# network 192.168.0.0 255.255.255.0
115+
router(dhcp-config)# default-router 192.168.0.254
116+
router(dhcp-config)# ip dhcp pool vm-vlan100
117+
router(dhcp-config)# network 192.168.100.0 255.255.255.0
118+
router(dhcp-config)# default-router 192.168.100.254
119+
router(dhcp-config)# ip dhcp pool vm-vlan200
120+
router(dhcp-config)# network 192.168.200.0 255.255.255.0
121+
router(dhcp-config)# default-router 192.168.200.254
122+
router(config-if)# interface ethernet0/2
123+
router(config-if)# ip address 192.168.0.254 255.255.255.0
124+
router(config-if)# no shutdown
125+
router(config-subif)# interface ethernet0/2.100
126+
router(config-subif)# encapsulation dot1q 100
127+
router(config-subif)# ip address 192.168.100.254 255.255.255.0
128+
router(config-subif)# interface ethernet0/2.200
129+
router(config-subif)# encapsulation dot1q 200
130+
router(config-subif)# ip address 192.168.200.254 255.255.255.0
131+
router(config-subif)# end
132+
router# copy running-config startup-config
133+
```
134+
135+
## Network Access to VMs from Different VLANs
136+
137+
1. **Network connection between VM networks**:
138+
139+
The router configuration above uses the [`A router on a stick`](https://www.grandmetric.com/knowledge-base/design_and_configure/router-on-a-stick-approach-cisco-configuration/) technology to allow VMs among untagged network, VLAN 100, and VLAN 200 to communicate with each other. Thus, it's not required to add any more configurations to the router.
140+
141+
2. **Network connection between VM networks and the management network**:
142+
143+
A feasible method to ensure network connectivity between VM networks and the management network is to manually add static routes. The following commands add static routes on the router to allow VMs in the untagged network, VLAN 100, and VLAN 200 to access the management network.
144+
145+
```
146+
router(config)# config terminal
147+
router(config)# ip route 10.10.0.0 255.255.255.0 ethernet0/1
148+
router(config)# ip route 192.168.0.0 255.255.255.0 ethernet0/2
149+
router(config)# ip route 192.168.100.0 255.255.255.0 ethernet0/2
150+
router(config)# ip route 192.168.200.0 255.255.255.0 ethernet0/2
151+
router(config)# end
152+
```
153+
154+
The route table would be like this:
155+
156+
```
157+
Router#show ip route
158+
Codes: C - connected, S - static, R - RIP, M - mobile, B - BGP
159+
D - EIGRP, EX - EIGRP external, O - OSPF, IA - OSPF inter area
160+
N1 - OSPF NSSA external type 1, N2 - OSPF NSSA external type 2
161+
E1 - OSPF external type 1, E2 - OSPF external type 2
162+
i - IS-IS, su - IS-IS summary, L1 - IS-IS level-1, L2 - IS-IS level-2
163+
ia - IS-IS inter area, * - candidate default, U - per-user static route
164+
o - ODR, P - periodic downloaded static route
165+
166+
Gateway of last resort is not set
167+
168+
C 192.168.200.0/24 is directly connected, Ethernet0/2.200
169+
10.0.0.0/24 is subnetted, 1 subnets
170+
C 10.10.0.0 is directly connected, Ethernet0/1
171+
C 192.168.0.0/24 is directly connected, Ethernet0/2
172+
C 192.168.100.0/24 is directly connected, Ethernet0/2.100
173+
```
174+
175+
## Access Harvester Load Balancers from Different VLANs
176+
The Harvester load balancer is divided into two types: VM load balancer and guest Kubernetes cluster load balancer.
177+
178+
1. The load balancer IP of the VM load balancer is only exposed within the same network as the Harvester hosts, or in other words, the management network. To access the VM load balancer from outside the network, you have to prove the route from outside client to the management network. For example, if the VM load balancer `lb1` has obtained its load balancer IP via DHCP and you want to access it from the VM `VM5`, you can add the following static route.
179+
180+
```
181+
router(config)# ip route 10.10.0.0 255.255.255.0 ethernet0/1
182+
router(config)# ip route 192.168.0.0 255.255.255.0 ethernet0/2
183+
```
184+
185+
2. The load balancer IP of the guest Kubernetes cluster load balancer is exposed within the VM network. In the diagram above, the guest cluster `demo` is within the VM network `vlan200` because the VMs consisting of the guest cluster are in the `vlan200`. Thus, the guest Kubernetes cluster load balancer `lb2` is exposed within the VM network `vlan200`. There are three scenarios to explain how to access `lb2` if it has obtained the load balancer IP via DHCP:
186+
- You can access it from the VM `VM3` and `VM4` directly because they are in the `vlan200`.
187+
- You can also access it from the VMs in other VM network directly because of the `A router on a stick` configuration.
188+
- You can access it from the Harvester hosts, or in other words, the management network by adding the following static routes on the router.
189+
190+
```
191+
router(config)# ip route 10.10.0.0 255.255.255.0 ethernet0/1
192+
router(config)# ip route 192.168.200.0 255.255.255.0 ethernet0/2
193+
```
194+
195+
196+
:::note
197+
198+
Except for the static routes above, you can also use dynamic routing protocols such as RIP, BGP, OSPF, and ISIS according to your network planning and requirements.
199+
200+
:::

‎docs/networking/deep-dive.md

Lines changed: 17 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -113,6 +113,23 @@ External network devices typically refer to switches and DHCP servers. With a cl
113113
| mode 5(balance-tlb) | none |
114114
| mode 6(balance-alb) | none |
115115

116+
For example, if the bond mode is 802.3ad, you need to configure LACP on the switch. The following is an example of LACP configuration on Cisco switch:
117+
```
118+
interface port-channel1
119+
switchport trunk encapsulation dot1q
120+
switchport mode trunk
121+
122+
interface g0/25
123+
switchport trunk encapsulation dot1q
124+
switchport mode trunk
125+
channel-group 1 mode active
126+
127+
interface g0/27
128+
switchport trunk encapsulation dot1q
129+
switchport mode trunk
130+
channel-group 1 mode active
131+
```
132+
116133
- If you want VMs in a VLAN to be able to obtain IP addresses through the DHCP protocol, configure an IP pool for that VLAN in the DHCP server.
117134
118135

‎static/img/v1.2/networking/best-practice.png

199 KB
Loading

0 commit comments

Comments
 (0)