Clustering

keuss-server can be run as a cluster out of the box with no extra configuration; all the needed bits are provided by keuss. All that is needed is, each node uses the same configuration so they end up with queues backed by the same mongodb or redis instance/cluster. Also, exchanges are copied and kept aligned on all cluster nodes (each node runs its own instances of the push-pop loops inside each exchange)

The clustering support improves if a Signaller other than local is used (and all nodes use the same configuration for it), cause insertions and rollbacks in queues will be known to all the cluster.

Cluster support improves even further if a Stats object other than mem is used (and all nodes use the same configuration for it): this would provide cluster-wide statistics and metrics.

Each exchange publishes its events on the signaller of its source queue; therefore, exchanges with source queues using a local signaller will not be propagated to the rest of the cluster