Green Cloud Vps Hong Kong’s High Availability Design And Multi-node Disaster Recovery Deployment Experience Sharing

1.

Background and Objectives

- Goal: Provide 99.99% availability and rapid fault recovery capabilities for businesses in Greater China.
- Node selection: Mainly using Hong Kong (HK) nodes, supplemented by Singapore (SG) and Tokyo (JP) as off-site backup.
- Business type: For e-commerce and API services, it is necessary to ensure session stability, database consistency and static resource acceleration.
- Constraints: single node bandwidth limit, DDoS attack risk, cross-border delay and compliance requirements.
- Success indicators: RTO ≤ 5min (automatic switching), RPO ≤ 1min (asynchronous/semi-synchronous replication strategy).

2.

High availability design principles

- Redundancy: at least three active or two active and one standby to avoid single points of failure (including links and power).
- Stateless: The application layer should be as stateless as possible, and the Session should exist in a Redis cluster or use JWT for horizontal expansion.
- Load balancing: The front end uses LVS/HAProxy/Nginx for layer seven or layer four balancing, combined with DNS low TTL.
- Health check: Actively detect health every 30 seconds. If failed three times in a row, it will trigger offline and notify the automated script.
- Data consistency: The master DB adopts master-slave semi-synchronization, separation of reading and writing, and synchronous backup of key transactions to reduce RPO.

3.

Key points of multi-node disaster recovery deployment

- Architecture model: Active-Active in the region + Active-Passive in remote areas is recommended to ensure a balance between reading, writing and latency.
- Data replication: MySQL semi-synchronous replication + binlog asynchronous archiving to object storage for secondary backup.
- DNS switching: Use authoritative DNS + health detection (TTL 30s). If necessary, manually force switching and notify CDN to refresh.
- BGP and Anycast: External traffic is accessed through Anycast or multi-line BGP to reduce the risk of single-link congestion.
- Capacity reservation: Each node reserves at least 30%-50% redundant bandwidth to cope with sudden traffic and DDoS surges.

4.

CDN and DDoS protection integration strategy

- Multi-layer protection: Connect to a third-party CDN for static acceleration and absorb normal attack traffic.
- WAF rules: Deploy WAF and RAT, SQL injection, CC protection rules at the edge and update them regularly.
- Anti-D threshold: Set the cleaning threshold based on historical peak values. For example, if the peak value is 5Gbps, the default cleaning threshold is 10Gbps.
- Black hole and cleaning: Automation first triggers traffic cleaning when there is an abnormality, and then uses the black hole strategy and alerts for manual intervention.
- Logs and traceability: Keep edge and origin site traffic logs for at least 30 days to facilitate attack analysis and legal evidence collection.

5.

Monitoring, automation and backup strategies

- Indicator collection: Prometheus collects host/application/database indicators, and Grafana displays the SLA panel.
- Alarm strategy: Severity classification (P0~P3), P0 automatically triggers the failover script and notifies by phone/text message.
- Configuration management: Use Ansible/terraform to manage images and networks to ensure repeatable deployment.
- Backup frequency: Full snapshot is taken once a day, and incremental logs are synchronized to the object storage every 1 minute (RPO ≤ 1min).
- Recovery drill: Conduct cross-region drills every quarter to verify DNS switching, database master-slave switching and traffic reflow.

6.

Real cases and server configuration examples

- Case overview: An e-commerce platform encountered sustained high concurrency and short-term DDoS during Double 11. It adopted Green Cloud HK node as the main site and enabled SG/JP as backup. It successfully completed the traffic switch within 10 minutes and controlled the loss within an acceptable range.
- Attack peak: peak traffic 5.2Gbps, real business traffic after cleaning is 3.8Gbps; protection threshold is set to 10Gbps.
- Availability results: After the switch, the system maintained order writing consistency, and the final statistical availability was 99.992%.
- Deployment instructions: The automated script completes the backend upstream switching within 90s after the monitoring is triggered and completes the CDN back-to-source verification within 180s.
- Sample configuration table (node specifications and roles):

Node	Role	CPU	Memory	Disk	Bandwidth
HK-node-1	Main application + load balancing	4 vCPU	8 GB	100 GB SSD	1 Gbps
SG-node-1	Standby application + read-only DB	4 vCPU	8 GB	100 GB SSD	1 Gbps
JP-node-1	Backup Snapshot and Monitoring	2 vCPU	4 GB	80 GB SSD	500 Mbps

Summary: High availability and multi-node disaster recovery based on Green Cloud VPS Hong Kong nodes require collaborative design in architecture, data replication, traffic protection and automated operation and maintenance. Through low TTL DNS, CDN edge cleaning, database semi-synchronous replication and regular drills, RTO can be controlled within minutes in emergencies and RPO can be guaranteed to be acceptable.