Language: Deutsch · English
Cluster service
Run several Nuclos servers as a cluster – master/slave, load balancer, sticky sessions and the cluster server view.
HOW-TO ADMINISTRATOR UPDATED: JUL 2026 APPLIES TO NUCLOS 4.2026.X
On this page
Menu Administration → Cluster service. From Nuclos 4.43 several Nuclos servers can run as a cluster – for load balancing and failover. All nodes use the same database and the same document directory. A dispatcher/load balancer sits between clients and servers (not part of the delivery).
Setup (in the installer)
Cluster operation is enabled/disabled during installation in the server configuration. Parameters:
| Parameter | Description |
|---|---|
| Node hostname | Hostname at which the node is directly reachable from other nodes (improves security, e.g. master/maintenance checks). |
| Node type = master | Runs jobs, distributes notifications and performs the DB auto-setup on update. Only one master may run at a time. |
| Node type = slave | Serves load balancing. |
| Load balancer protocol/hostname/port/context path | Optional – only if different from the Nuclos server. |
Load balancer
Since Nuclos ships no balancer, its parameters must be set: the REST service embeds URLs to other services in results – these should point to the balancer. Therefore the load balancer parameters take precedence for URL generation in a cluster.
Sticky sessions required
Sticky sessions are mandatory: a client's session must always talk to the same node (also for SSO tokens). If a node fails, its sessions become invalid – re-login required.
Cluster server view
In the web and rich client (menu Administration) you can view the list of registered cluster servers (read-only): node status, exchanged messages (max. 24 h) and a live status query of all servers via REST (issued by the current server, marked as tester).
Limitation
In cluster operation the Lucene indexer is disabled automatically by the installer.
Related pages
Überblick
Inhalte