Which Alerts Should Database Monitoring Have? The DBA Alert Matrix

A DBA alert matrix is a document that defines, in a single table, which metric raises an alert at which threshold in your database monitoring, how severe that alert is and who does what when it fires. A good alert matrix prevents two failures at once: a critical problem that never raises an alert, and a team that misses the real problem among trivial alerts.

This guide explains how to build an alert matrix for SQL Server and PostgreSQL environments. The suggested thresholds are a starting point; tune them to your own system’s normal behavior (its baseline).

What Makes a Good Alert Matrix

Severity Levels

The Alert Matrix: Common Metrics

The table below lists the core alerts that apply to both SQL Server and PostgreSQL. The thresholds are starting suggestions.

SQL Server-Specific Alerts

SQL Server Agent alerts for errors 823, 824 and 825 take only a few lines. The example below creates the alerts; for notifications you need to define an operator and attach it with sp_add_notification:

USE msdb;
EXEC dbo.sp_add_alert @name = N'Error 823 - I/O error',
     @message_id = 823, @severity = 0, @include_event_description_in = 1;
EXEC dbo.sp_add_alert @name = N'Error 824 - Logical consistency error',
     @message_id = 824, @severity = 0, @include_event_description_in = 1;
EXEC dbo.sp_add_alert @name = N'Error 825 - Read retry',
     @message_id = 825, @severity = 0, @include_event_description_in = 1;
EXEC dbo.sp_add_alert @name = N'Severity 024 - Fatal hardware error',
     @severity = 24, @include_event_description_in = 1;

We cover which alerts to enable and how to set up operators in detail in our SQL Server alerts guide.

PostgreSQL-Specific Alerts

For example, these queries for transaction ID age and inactive slots can be added to your monitoring system:

-- Transaction ID age as a percentage of autovacuum_freeze_max_age
SELECT datname,
       age(datfrozenxid) AS xid_age,
       round(100.0 * age(datfrozenxid) / current_setting('autovacuum_freeze_max_age')::int, 1) AS pct_of_freeze_max
FROM pg_database
ORDER BY xid_age DESC;
-- Inactive replication slots and the WAL they retain
SELECT slot_name, active,
       pg_size_pretty(pg_wal_lsn_diff(pg_current_wal_lsn(), restart_lsn)) AS retained_wal
FROM pg_replication_slots
WHERE NOT active;

Preventing Alert Fatigue

Too many alerts can be more dangerous than none: the team gets used to the constant stream of notifications and ignores the critical one too. To prevent this:

• Add a duration condition: use conditions like “for 15 minutes” instead of momentary CPU spikes.
• Group repeats: don’t send a new notification every minute for the same problem.
• Define maintenance windows: silence expected alerts during planned work.
• Review monthly: report how often each alert fired and how many times it needed real action.

An Alert Matrix Is a Document, Not a Tool

Which monitoring tool you use is a decision that comes after the alert matrix. First define what is watched, at which threshold and by whom; then choose a tool that can implement that matrix. You’ll find daily, weekly and monthly checks in our SQL Server monitoring checklist, and the first steps during an outage in our first 30 minutes runbook.

24/7 Monitoring with Aryasoft

Under Aryasoft’s SLA model, monitoring runs 24/7/365: alerts open tickets automatically, an engineer from the on-call team resolves them and every incident appears in the monthly report. The team is made up of an account manager, a technical team and an on-call team. First response is 30 minutes during business hours and 2 hours after hours. We build the alert matrix with you on day one, based on your systems’ normal behavior. Details are on our managed database services page.

Frequently Asked Questions

Sources

The technical details in this guide were checked against the following official documentation in November 2026.

1. Microsoft Learn: MSSQLSERVER_825
2. Microsoft Learn: MSSQLSERVER_824
3. Microsoft Learn: Database engine error severities
4. Microsoft Learn: sp_add_alert
5. PostgreSQL documentation: Routine Vacuuming
6. PostgreSQL documentation: pg_replication_slots