facebook/rocksdb

Choose this fork if you need TiKV-oriented RocksDB behavior and can tolerate a large upstream gap. Choose upstream if you want the broadest feature set, fastest access to recent fixes, and lower long-term merge risk.

Prefer this fork only if you specifically need its legacy customizations and are willing to own a large rebase burden. For most adopters, upstream RocksDB is the safer choice because this fork is materially outdated and highly divergent.

Choose this fork only if SPDK integration is the requirement; for most adopters, upstream RocksDB is the safer default because this fork is heavily diverged and stale.

Choose this fork only if its NVM/SSD backend work is the reason you need it. For most adopters, upstream RocksDB is the safer default because this fork is materially stale, heavily divergent, and likely missing years of fixes and developer-facing improvements.

Choose this fork only if rDSN-based replication is the product goal. For most adopters, upstream RocksDB is the better base because this fork is very stale, highly divergent, and likely missing many years of upstream fixes and features.

Prefer upstream unless you specifically need the fork's local patches and are prepared to own long-term maintenance. This fork looks too stale and too divergent for most adopters.

Choose this fork only if you need the Ceph-specific patch set and are willing to own a long-lived divergence from upstream. If you want current RocksDB features, fixes, and APIs, upstream is the better default.

Choose this fork only if rDSN-based replication is the goal and you can absorb a stale, highly divergent codebase. For most adopters, upstream RocksDB is the safer choice because this fork is materially outdated and specialized.

Choose this fork only if its backup/restore and read-only DB behavior matches a real requirement you cannot get from upstream. For most adopters, upstream RocksDB is the safer default because this fork is materially behind and appears stale.