How Do I Restore a Deleted VPC Endpoint for a Cluster Connected Over a Private Network?

Symptom

The VPC endpoint is deleted by mistake in such a cluster, and the cluster becomes abnormal.

Procedure

The IP address of the VPC endpoint has been configured in proxy-agent. Therefore, you need to specify an available IP address when creating a VPC endpoint.

1. Log in to the VPC Endpoint console to check whether the VPC endpoint in the region where UCS is located is deleted. If yes, go to the next step.
2. Log in to the master node of the abnormal cluster.
3. Query the IP address configured in proxy-agent.

   kubectl get deploy -n kube-system proxy-agent -oyaml | grep -A3 hostAliases

   Command output:

   hostAliases:
   - hostnames:
     - proxyurl.ucs.myhuaweicloud.com
     ip: 10.0.0.182

4. Create a VPC endpoint in the region where UCS is located, use the preceding IP address, and click View In-Use IP Address to ensure that this IP address is not used by any other VPC endpoint. If the IP address is in use, change the proxy-agent configuration in the cluster. For details, see Changing proxy-agent Configurations.
   Figure 1 Buying a VPC endpoint (with manually specified private IP address)
   

5. Click Next and then Submit to create a VPC endpoint again.
6. Wait for 1 to 3 minutes, go back to the UCS console, and refresh the cluster status.

Changing proxy-agent Configurations

1. Create a VPC endpoint in the region where UCS is located.
   Figure 2 Buying a VPC endpoint (with automatically assigned private IP address)
   

2. Click the ID of the newly created VPC endpoint to view the automatically allocated node IP address.
   Figure 3 Viewing VPC endpoint details
   

3. Log in to the master node of the abnormal cluster.
4. Change the IP address configured in proxy-agent.

   kubectl edit deploy -n kube-system proxy-agent

   Change the IP address in the hostAliases field.

   hostAliases:
   - hostnames:
     - proxyurl.ucs.myhuaweicloud.com
     ip: 10.0.0.122

   Press Esc, enter :wq, and press Enter.

5. Wait for 1 to 3 minutes, go back to the UCS console, and refresh the cluster status.