Database Monitoring
Read-only checks in a dedicated health database
SBN Monitor watches a database server through a small set of read-only health checks installed in a dedicated health database on that server. The checks live in the database, so adding or refining one is a database change on its own schedule. The tool calls a check, reads back a fixed result of named readings, and turns those into sensors. The checks never write to or alter the monitored data.
A restricted monitor login
The monitor connects with a dedicated login that holds only permission to execute the health checks. This retires the old practice of handing privileged database credentials to the monitoring path. The connection uses the server's own configured, hardened authentication. The health database reports a version the tool confirms on connect.
What the checks cover
The health checks report the server's state across these areas:
Area | What it reports |
|---|---|
Version and heartbeat | the server is up and answering, and which build it runs |
Databases | the list of databases present and each one's state |
Space | data space used and free, per database, in both percentage and absolute size |
Log space | transaction-log space used and free, per database |
Connections | user connections in use against the configured ceiling |
Long-running queries | the longest-running work in progress |
Queue depth | how far behind a monitored processing queue is |
Deadlocks | deadlocks seen |
Replication | replication partition space and whether replication is flowing |
Each reading becomes a sensor with its own thresholds, timing, and pause state, tuned centrally.
Per-database sensor discovery
The per-database checks - data space and log space especially - are materialized from what the server actually holds. The health database enumerates the databases present, and the tool creates a sensor per database from that list automatically. Adding or dropping a database never means editing sensor configuration; each database keeps its own thresholds and pause state. Central configuration carries only the exceptions - a per-database threshold override, or an expected-database list so a server that deliberately holds a subset never alarms on a missing database.
Supported database servers
One database sensor type covers every supported platform, with the right driver and health-check source chosen per server:
- Sybase ASE
- Microsoft SQL Server
The readings, thresholds, and routing are the same across platforms, though the checks underneath are written for each server's dialect.