PostgreSQL monitoring
PostgreSQL monitoring without agents
Scops monitors every PostgreSQL instance from one place, next to your Oracle and SQL Server databases. Nothing is installed on the server, one role in pg_monitor is enough, and 29 Radar detectors report wraparound, a held xmin horizon or a slot that keeps WAL while there is still time to act.
What Scops watches on PostgreSQL
- Activity: active backends by wait event type, with their parallel workers, against the usual level of that hour.
- Statements: the busiest statements by
query_id, each with its text, counters and history. A plan regression is flagged. - Sessions and locks: every session now, each lock chain with its root, sessions idle in a transaction and the oldest open transaction. A prepared transaction at the root of a chain is named too.
- Long commands: VACUUM, ANALYZE, CREATE INDEX, CLUSTER, base backups and COPY, with their phase and progress.
- Deadlocks: each one with its processes, what they waited for and the statement the server ended.
- Vacuum and wraparound: transaction ID age per database, the tables autovacuum does not keep up with, and what holds the xmin horizon.
- Replication and WAL: standbys and their lag, replication slots and the WAL they keep, and the archiver.
- Space: databases, tablespaces, WAL, temporary files, tables and indexes, with their growth.
Radar Engine for PostgreSQL
Radar Engine catches the problems that are hardest to see before they break an instance. For a problem that already happened, it is the strongest way to reach the root cause.
On PostgreSQL it runs 29 detectors:
| Group | Detectors |
|---|---|
| Load | Blocking pile-up, wait wave, load surge, throughput collapse, synchronous commit wait, deadlock burst, rollback surge, WAL surge |
| Resources | Connection limit, CPU saturation, memory running out, I/O latency |
| SQL | One statement taking over, plan regression, temp file spill |
| Space | Disk filling, transaction ID wraparound, sequence running out, slot holding WAL |
| Availability | Standby not streaming, replication lag, role change, restart, WAL archiver failing, corruption |
| State | Vacuum falling behind, xmin horizon held, configuration change, record hour |
A finding names its cause. Wraparound is reported with the reason the age does not fall: a session, a prepared transaction or a replication slot that holds the xmin horizon, autovacuum turned off, or every autovacuum worker busy.
Agentless, with one role
Scops connects to the instance at its host and port like any client, with one role that is a member of pg_monitor. It is not a superuser. Scops reads every database of the instance through that connection, and nothing is installed on the server.
Sessions, waits, locks and Radar work from the first minute. Where pg_stat_statements is loaded, Scops adds the statistics of every statement. An optional OS credential over SSH adds the CPU, memory and disk of the host and the server log, where PostgreSQL writes the detail of a deadlock. The connection test says what it found and writes the steps for the rest.
What it monitors
- PostgreSQL 16, 17 and 18
- Primaries and standbys, with streaming replication and replication slots
- Every database of an instance through one registration
One platform for Oracle, SQL Server and PostgreSQL
The same installation monitors Oracle and SQL Server. Each engine has its own Dashboard, Live Activity, Insights, Metrics, Health Check, Sizing and Radar detectors, in its own terms, and all of them share one Databases page, one alert log and one set of reports. See database monitoring with Scops.
Questions
Does Scops install an agent on the PostgreSQL server?
No. Scops connects to the instance like any client and installs nothing on the server.
Which role does Scops need on PostgreSQL?
One role that is a member of pg_monitor, with CONNECT on the database the connection opens in. It is not a superuser. The connection test writes the script for anything that is missing.
Does Scops need pg_stat_statements?
Scops samples the sessions itself, so activity, waits, locks and Radar Engine work from the first minute. Where pg_stat_statements is loaded, Scops adds the statistics of every statement.
How does Scops warn of transaction ID wraparound?
Radar Engine follows the transaction ID and multixact age of every database and how fast it rises, and reports it long before PostgreSQL stops accepting writes. The finding says why the age does not fall: what holds the xmin horizon, autovacuum turned off, or every worker busy.
Which PostgreSQL versions does Scops monitor?
PostgreSQL 16, 17 and 18, primaries and standbys.
What does PostgreSQL monitoring cost with Scops?
Community is free for 2 PostgreSQL instances, with Radar Engine, alerts and reports. Professional is $249 per instance a year, each further one costs 3% less, down to $199. An instance counts once, however many databases it holds.
Can the same installation monitor Oracle and SQL Server?
Yes. Oracle, SQL Server and PostgreSQL are monitored from one Scops server, on one Databases page.
Start
Open the live demo, or download Scops. It is free for 2 PostgreSQL instances. To connect the first one, see Add a database and system requirements.