Data disaster recovery

Solution selection

Elasticsearch offers two solutions for off-site disaster recovery:

OSS snapshot backup and restore: Backs up index data to Alibaba Cloud Object Storage Service (OSS). The first snapshot is a full backup, and subsequent snapshots are incremental. You can restore data to another Elasticsearch instance through a cross-cluster OSS repository. For more information, see Back up and restore data by using a cross-cluster OSS repository.
Cross-cluster replication (CCR): Asynchronously and incrementally replicates writable indexes from a leader cluster to one or more follower clusters, achieving near-real-time sync. Ideal for disaster recovery with strict RPO and RTO requirements. For more information, see Replicate data across clusters by using CCR.

Solution	Scenarios	RPO	RTO	Major limitations
OSS snapshot	Periodic backup and restore of large-scale data (GB to PB).	Hours to days (depends on snapshot interval).	Several hours (depends on data volume and shard recovery).	No continuous sync. Downtime may be required during restoration.
CCR	Off-site disaster recovery, read/write splitting, and proximity-based access.	Near-zero (seconds).	Seconds to minutes.	Follower indexes are read-only. Requires identical mappings and shard counts.

For off-site disaster recovery that requires low RPO and real-time availability, CCR is recommended:

Synchronizes data in seconds, minimizing data loss.
If the primary cluster fails, you can switch traffic to the follower cluster immediately without waiting for snapshot restoration.
Higher initial cost, but more cost-effective long-term by preventing business losses from data unavailability.