> For AI agents: the complete documentation index is available at /llms.txt, the full documentation bundle is available at /llms-full.txt.

# Supported databases

Every database Ivory supports is a plugin with two halves:

- **The cluster side** — how Ivory finds the members of a cluster, tells the leader from the replicas, measures
  replication lag and runs operations like a switchover or failover. In Ivory's terms this is the cluster's
  [Keeper](/docs/overview.md#what-is-a-keeper): an HA tool such as Patroni, or the engine's own built-in replication.
  It is reached on the node's **Keeper port**.
- **The database side** — the query console, saved queries and dashboards, reached on the node's
  **Database port**. For engines that manage their own replication, both ports are usually the same.

Because of that split, every cluster looks the same in Ivory — one [overview](/docs/overview.md), one set of
operation buttons, one console — and each page below describes what that engine fills in.

| Database                                           | Cluster                                   | Stage  |
| -------------------------------------------------- | ----------------------------------------- | ------ |
| [PostgreSQL + Patroni](/docs/databases/patroni.md) | High-availability cluster with Patroni    | STABLE |
| [PostgreSQL](/docs/databases/postgres.md)          | Primary and standby streaming replication | STABLE |
| [etcd](/docs/databases/etcd.md)                    | Raft cluster                              | BETA   |
| [Redis](/docs/databases/redis.md)                  | Master and replica replication            | BETA   |
| [ClickHouse](/docs/databases/clickhouse.md)        | Replicated cluster                        | BETA   |
| [MongoDB](/docs/databases/mongodb.md)              | Replica set                               | BETA   |
| [ZooKeeper](/docs/databases/zookeeper.md)          | Ensemble                                  | BETA   |

**STABLE** is what Ivory has run against for a long time; **BETA** works end to end but has had less time in
the field; new plugins start at **ALPHA**.

## Operations

An operation a cluster cannot perform is simply not offered in the UI. ⏱ marks one that can also be
scheduled for a later time.

| Cluster    | Operations                                                                                    |
| ---------- | --------------------------------------------------------------------------------------------- |
| Patroni    | Switchover ⏱, failover, restart ⏱, reinitialise, reload, pause / resume, view and edit config |
| PostgreSQL | Failover (promote a standby), reload, view config                                             |
| etcd       | Switchover (move the leader)                                                                  |
| Redis      | Failover (promote a replica), view config                                                     |
| ClickHouse | Reload, view config                                                                           |
| MongoDB    | Failover (step the primary down), view and edit config                                        |
| ZooKeeper  | View config                                                                                   |

## Health and replication lag

Ivory warns about a replica whose lag passes its threshold. The threshold is Ivory's own default for that
engine, not a setting of the engine itself.

| Cluster    | Replication | Lag is measured in                      | Warning threshold |
| ---------- | ----------- | --------------------------------------- | ----------------- |
| Patroni    | One leader  | Bytes behind the leader                 | 50 MiB            |
| PostgreSQL | One leader  | Bytes of WAL not yet replayed           | 50 MiB            |
| etcd       | One leader  | Raft entries behind the leader          | 1000 entries      |
| Redis      | One leader  | Seconds since the master was last heard | 30 s              |
| ClickHouse | Every node  | Seconds of replication delay            | 30 s              |
| MongoDB    | One leader  | Seconds behind the primary              | 30 s              |
| ZooKeeper  | One leader  | Not measured                            | —                 |

## Query console

| Cluster    | Console                         | Browse                     | Running queries    | Dashboard |
| ---------- | ------------------------------- | -------------------------- | ------------------ | --------- |
| Patroni    | SQL                             | Databases, schemas, tables | List, cancel, kill | ✅         |
| PostgreSQL | SQL                             | Databases, schemas, tables | List, cancel, kill | ✅         |
| etcd       | etcdctl-style commands          | —                          | —                  |           |
| Redis      | Redis commands                  | Numbered databases         | List, kill clients |           |
| ClickHouse | SQL                             | Databases, tables          | —                  | ✅         |
| MongoDB    | Shell-style `collection.verb()` | Databases, collections     | List, kill         |           |
| ZooKeeper  | znode commands                  | —                          | —                  |           |

Every engine ships a set of saved queries for monitoring and diagnostics; see [Node → Database](/docs/node.md#database).

## Adding your database

Every engine is a self-contained plugin, so supporting a new one needs no change to Ivory's core. The
repository ships an AI agent for exactly this — ask Claude Code to use `keeper-database-plugin` for your engine
and it writes both halves, their tests and the UI wiring by the project's checklist. Then open a pull request;
see [Contributing](/docs/community/contributing.md).