Lite Server Hot Standby Nodes

Scenario

To enable Lite Server hot standby, create custom Kubernetes clusters. These clusters mark specific nodes as hot standbys by applying taints, so that service pods will not be scheduled to these nodes.

Constraints

Prepare required hot standby nodes by referring to the following table.

**Table 1** Required hot standby nodes
Resource Type/Required Nodes	10–49	50–99	100–249	250–499	500–749	750–1,000	> 1,000
Snt9A	1	2	3	5	7	10	12
Snt9b	1	2	3	4	5	6	10
GP Ant8	1	2	3	5	6	8	12
GP Vnt1	1	2	3	5	8	10	12

Example 1:

If you have purchased six Snt9b nodes, which is less than 10. According to the preceding table, you do not need to prepare hot standby nodes.

Example 2:

If you have purchased 600 Snt9b nodes, which is between 500 and 749. According to the preceding table, you need to prepare five more nodes as hot standbys. In this case, you need to purchase 605 nodes in total.

Prerequisites

You have created a Kubernetes cluster using Lite Server.

Hot Standby Replacement

If a hardware fault occurs on a service node in the cluster and hot standby replacement is required, back up data first, and then taint the faulty node and remove the hot standby node's taint.

Data backup
Use rsync, a powerful file synchronization tool that supports local and remote synchronization, to flexibly and efficiently back up data.
```
rsync -avz -e ssh /source/ user@remote:/destination/
```
Back up the faulty node file to the hot standby node. The backup.txt file is used as an example.

Complete the backup using SSH. Then, you can view the file on the new hot standby node.
Hot standby node replacement
Taint the faulty node.
```
kubectl taint nodes <node-name> dedicated=ops:NoSchedule
```
- <node-name>: Replace it with the actual node name.
- dedicated=ops: key–value pair of the taint.
- NoSchedule: taint effect. This indicates that kube-scheduler will not schedule pods to the node.
Check whether the node is tainted.
```
kubectl describe node <node-name> | grep Taints
```
The node is labeled as unschedulable.

Remove the taint from the hot standby node.
```
kubectl taint node <node-name> dedicated=ops:NoSchedule-
```
Ensure that the taint is removed.
```
kubectl describe node <node-name> | grep Taints
```
The hot standby node replacement is complete. Repair the faulty node.