Skip to main content
Guidelines for storing, backing up, and restoring persistent data in production.

​
Volume layout (defaults)

ServiceDefault path
TiKV/data
PD/data
Kafka/var/lib/kafka/data
Redis/data
MongoDB/data/db
Use SSD or NVMe-backed storage for all stateful volumes.

​
Backup strategy

  • TiDB: daily backups to secure, off-cluster storage
  • Kafka: weekly segment-level backups
  • Redis: RDB snapshots every 6 hours (cache is non-authoritative)
  • Backup validation: monthly restore and integrity verification tests

​
Disaster recovery

  • Validate full restore procedures at least once per quarter
  • Maintain a minimum of three geographically isolated backup copies
  • Run staged DR simulations such as warm-standby restoration and full cluster rehydration