rocksdb-checkpoint · shipping · also CreateCheckpoint

RocksDB checkpoint

A RocksDB checkpoint is a captured database directory that can be opened separately from the source database.

Related Observatory article: RocksDB checkpoint on hesela.com

On the same filesystem, SST and blob files use hard links; other required files are copied. Files are copied when crossing filesystems. By default, log_size_for_flush is zero and all column families are flushed. Raising it can omit unflushed no-WAL writes. [1]

A different pathname is not a separate failure domain. A checkpoint sharing the source filesystem cannot by itself survive loss of that filesystem. This follows from the shared-file layout, not a measured recovery probability.

Relations

related
rocksdb-read-snapshot, rocksdb-backup-engine, hard-link, rocksdb-column-family, write-ahead-log, backup, restore-test

Citations

  1. RocksDB v10.10.1. checkpoint.h: file capture and flush threshold. (accessed 2026-10-17)

Further reading

  1. Hesela: RocksDB capture and recovery boundaries. (accessed 2026-10-17)
Sources & caveats

JSON: /data/terms/rocksdb-checkpoint.json