Glossary

TeselaDB's terms, with the same name across the documentation.

Each concept always has the same name in this documentation.

Interval [start, end) of the ordered key space, with its own Raft group. It is the unit of replication, placement and splitting.

Copy of a range on a node. It can be a voter or a learner.

Replica that proposes the entries of its range. With a lease it answers reads with no messages.

Metadata Raft group of the cluster: nodes, catalogue, descriptors and configuration.

Thread that solely owns its event loop, its Raft log, its LSM and its ranges.

Ordered store of a core with a memtable, SST files and compaction.

Immutable, sorted file of the LSM.

8 bytes of physical milliseconds and 4 of logical counter. It orders versions and reads.

Timestamp below which the leader of a range will accept no writes. It allows reads on followers.

Provisional version of a key written by a distributed transaction.

Prefix that spreads the entries of an index on a growing value (teselaBuckets).

How the cluster shows itself to drivers: mongos (the default) or replica set.

Observable difference with MongoDB, recorded and checked by the compatibility suite.

Tests that run the real drivers and mongosh against TeselaDB and compare with MongoDB.

Credential tdb_… tied to a user, with narrowed roles and an expiry.

How much MVCC history is kept: 24 hours by default (teselaHistoryRetention).