3-2-1 Golden Rule

3 copies of data (original plus 2 backups). 2 different storage media (local disk plus object storage/tape). 1 copy offsite (different geographic region against fire/earthquake/flood). For most 益阳 SMEs: Original on cloud server. Backup 1: vendor snapshot (same DC, different machine). Backup 2: scheduled script pushing to OSS (same city different DC or cross-region). Monthly cost: 10-30 RMB for typical 10-20GB site.

What to Back Up

Website files: source code, configs, uploaded images/attachments, static assets. Database: all dumps (MySQL dump, Redis RDB/AOF, MongoDB dump). Config files: Nginx/Apache configs, crontab, SSL certs. App configs: .env files containing passwords/API keys. Most commonly missed: config files and .env - losing them means manual reconstruction taking hours.

Backup Frequency

High-change (e-commerce/social/content platform): hourly or every 4-hour full DB plus real-time binlog sync. Low-change (corporate/display site): daily sufficient. Key moments: manual full backup before major events.

A 益阳 e-commerce company had operator accidentally truncate product table on Double 11 afternoon 3 PM. Fortunately hourly auto-backup existed - only lost last 45 minutes of orders (most recoverable via binlog). Actual loss under 2,000 RMB. Without that hourly backup: could have lost entire day orders (estimated GMV 800,000 RMB that day).

Restore Testing

Backups that cannot restore are useless. Quarterly: take latest backup, restore to test server, verify data integrity, record time taken, note issues found and fix backup script accordingly. One client backup script had gzip corruption bug undiscovered for months because nobody ever tested restoration.