[HCI-VN] Physical network cross-connection scenarios cause Virtual Network anomalies
Problem Description
- Referring to the current status of the previous fault case, the link aggregation method of the HCI all-in-one machine and the business network switch first chooses to load based on network port polling, which causes problems. An error occurs when deleting a router in Virtual Network. The solution to this problem is to use the MAC address for loading.
[color=rgb(0, 104, 189) !important]https://support.sangfor.com.cn/cases/read?product_id=33&category_id=17385
2. After the modification is completed according to the above case, problem 2 arises: the first ping packet of the customer's business network has a very large delay and is prone to packet loss.
Alarm Information
No warning information on the platform
Effective Troubleshooting Steps
1. Find the customer's network operation and maintenance personnel to confirm the network topology again, as shown below
2. Confirm that the core switch previously used link aggregation based on MAC address;
3. Confirm that the cross-switch link aggregation mode of the service network switches and core switches is the link aggregation mode based on the MAC address format;
4. Confirm that the cross-switch link aggregation mode of the hyper-converged appliance and the service network switch is a link aggregation mode based on the MAC address format;
5. Confirm that Virtual Network crosses the core switch and other network segments to perform ping packet tests without firewall devices blocking
6. When one of the network cables of the link aggregation of the service network is unplugged, the test service is normal.
Root Cause
- The virtual machine to the business network switch is a cross-switch link aggregation method, and load routing is based on MAC address.
2. The service network switch to the core switch is also a cross-switch link aggregation method, and load routing is performed based on MAC address
3. There is a certain probability that the two-segment interconnection routing process will lead to inconsistent routing results, which will cause the first packet loss phenomenon. In the actual environment, the first session will not be maintained for a long time, and the TCP session link will be easily disconnected after a while.
Solution
1. The server to the business network switch does not use cross-switch link aggregation, but uses a single business switch networking form
2. Continue to use the cross-switch link aggregation method, but select the active-standby mode for the routing method. Give priority to one of the service network switches.
Suggestions and Summary
The cross-switch link aggregation of multiple devices cross-connected will give priority to the active/standby mode of routing.
Original Link
https://support.sangfor.com.cn/cases/list?product_id=33&type=1&category_id=16247&isOpen=true