Observing a multi-node deployment previously meant querying each node's monitoring catalog separately. This release adds a federated catalog that answers a single query across the whole cluster, so an operator reads node health from one place.
Two-catalog architecture
- Cluster mirrors instance. A new cluster catalog mirrors every schema in the existing per-instance catalog — JVM, cache, machine and statistics — and prepends three identity columns, node id, node name and node address, to every result row.
- Parallel fan-out. A read query runs in parallel against all online nodes, the local one in process and remote ones over JMX, with a five-second timeout in tolerant mode so an offline node is skipped rather than failing the whole query.
Per-node activity
- Query counts and errors. Each node now tracks the number of fan-out queries it has answered successfully and its last connectivity error, cleared automatically once a health probe confirms the node is back online.
- Wired into the explorer. The cluster catalog is exposed as a second database under the runtime server node in the database explorer topology, alongside the existing per-instance catalog.
The per-instance catalog is unchanged and remains available; the cluster catalog is an additional, aggregated view for deployments that run more than one node.