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

# Deployment Topologies

JumpServer can run as a single all-in-one server, as a pair of nodes behind a load
balancer, or as a distributed deployment in which the database, cache, storage, and
search services run on dedicated hosts. This page compares the topologies and
lists the dependencies and prerequisites of each one.

## Topology comparison

| Topology          | Good for                                  | Key dependencies                                                        | Main failure domain             |
| ----------------- | ----------------------------------------- | ----------------------------------------------------------------------- | ------------------------------- |
| Single node       | Evaluation, small teams, lab environments | One Linux host with Docker                                              | The single host                 |
| Two-node HA       | Production with a modest footprint        | Load balancer, shared storage, external database and cache              | Load balancer or shared storage |
| Fully distributed | Large or regulated environments           | NFS or object storage, PostgreSQL, Redis, HAProxy, MinIO, Elasticsearch | Each dedicated service          |

## Single-node deployment

The default installation runs every service on one host with Docker Compose. Core,
the web frontend, and the protocol components run as containers, and a built-in
PostgreSQL and Redis are started next to them. Recordings, logs, and database
backups are written under `VOLUME_DIR`, `/data/jumpserver` by default.

This is the fastest way to evaluate JumpServer and is sufficient for small
production environments. The host is a single point of failure, so keep regular
[database and static file backups](https://www.jumpserver.com/docs/operations/backup-restore.md) and monitor disk
space under `VOLUME_DIR`.

## Two-node high availability

Two JumpServer nodes run the same services behind a load balancer. The nodes share
the Core data directory through NFS or use an object storage for recordings, and
they use an external PostgreSQL and Redis so that data is not tied to either node.

Key requirements:

- Install the first node, then reuse its `SECRET_KEY`, `BOOTSTRAP_TOKEN`, and
  `CHAT_AI_DELEGATION_SECRET` on the second node. Nodes that do not share these
  values cannot read the same data or register the same components.
- Point component registration at the cluster address instead of a single node.
- If a built-in database is used instead of an external service, bind the built-in
  database to each node's fixed HA address with `POSTGRESQL_EXPOSE_HOST` or
  `MYSQL_EXPOSE_HOST` and the matching `*_EXPOSE_PORT` settings.

## Distributed deployment

In a fully distributed deployment, the supporting services run on dedicated hosts:
NFS, PostgreSQL, Redis, HAProxy, MinIO, and Elasticsearch. Deploy them before the
JumpServer nodes, in this order:

1. NFS
2. PostgreSQL
3. Redis
4. JumpServer node 01
5. JumpServer node 02
6. HAProxy
7. MinIO (object storage)
8. Elasticsearch (command storage)

If the environment already provides an object storage service such as S3, Ceph,
Swift, or OSS, MinIO is not required. In production, prefer Ceph or a highly
available NFS instead of a single NFS server. A third-party load balancer can
replace HAProxy, as long as it supports session persistence and WebSocket
connections.

## Version requirements

The supporting services must meet these minimum versions:

| Service    | Requirement                                              |
| ---------- | -------------------------------------------------------- |
| PostgreSQL | 16 or later                                              |
| MariaDB    | 10.6 or later                                            |
| Redis      | 7.0 or later; Sentinel is supported, cluster mode is not |

Compression, TLS, and high-availability options for the database and cache are
documented in [Configuration](https://www.jumpserver.com/docs/operations/configuration.md).