Database monitoring

Know which query is slowing your database

Q: Which databases does Tracix support?

Health metrics for PostgreSQL, MySQL, SQL Server, Redis, MongoDB and Elasticsearch. Query-level monitoring, with top queries, samples, wait events and explain plans, covers PostgreSQL, MySQL and SQL Server.

Q: Do I install anything on the database server?

No. A collector that can reach the database connects over the network as a monitoring user, with TLS and client certificates if you turn them on. The queries it runs for monitoring are left out of your query views.

Q: What does PostgreSQL need for query monitoring?

The pg_stat_statements extension and the pg_monitor role for top queries and explain plans. Live query samples and wait events work without them.

Q: Are the recommendations written by AI?

No. They come from rules over your query telemetry, such as expensive scans, missing or unused indexes, regressions, lock contention and long-running queries. When an AI model is configured, Brain adds a short written summary.

Watch the health of PostgreSQL, MySQL, SQL Server, Redis, MongoDB and Elasticsearch through the collector. For PostgreSQL, MySQL and SQL Server, see the top queries, live samples with wait events, explain plans and database load, and which traced services call each database.

Start free Book a demo Read the docs

PostgreSQL · MySQL · SQL Server
Redis · MongoDB · Elasticsearch
Explain plans

Database monitoring5 databases · PostgreSQL, MySQL and RedisLast 15 minutes

DatabasesDatabase loadQueries

All databases · 214 normalized queriesSorted by total duration

Normalized queryCountAvgTotal

SELECT id, user_id, items FROM carts WHERE id = $1PostgreSQLorders-db · orders41.2k0.42 ms17.3 s

SELECT * FROM promotions WHERE active AND starts_at <= $1PostgreSQLorders-db · orders19.6k0.88 ms17.2 s

INSERT INTO ledger (charge_id, amount, currency) VALUES ($1, $2, $3)PostgreSQLpayments-db · payments2.1k3.80 ms8.0 s

UPDATE orders SET status = $1, updated_at = now() WHERE id = $2PostgreSQLorders-db · orders2.1k2.90 ms6.1 s

SELECT o.id, o.total FROM orders o JOIN customers c ON c.id = o.customer_id WHERE c.email = $1PostgreSQLorders-db · orders31218.40 ms5.7 s

SELECT name, price FROM products WHERE sku IN (?)MySQLcatalog-db · catalog8.4k0.61 ms5.1 s

The collector's own monitoring queries are hidden

How it works

From a monitoring user to the query that matters

The collector reads what the database already records about itself, so there is nothing to install on the database server.

01ConnectPick an engine, enter its endpoint and a monitoring user, and choose the collector policy that can reach it.
02CollectHealth metrics arrive every 30 seconds, with top queries, live samples and plans where the engine supports them.
03Find the loadDatabase load shows active sessions by wait group: CPU, IO, lock, commit and other.
04Drill inOpen a normalized query to see its latency, executions, waits, explain plan and recent samples.
05TuneRecommendations flag expensive scans, missing or unused indexes, regressions and lock contention.
06Trace it backUsage from traces shows which services call each database, from the spans they already send.

Queries and plans

Every normalized query, with its waits and its plan

Top queries are grouped by their normalized text, so a query that runs a million times with different values is one row. Open one to see where its time goes and the plan the database chose.

Live samples show each running query with its state, wait event and the user that ran it
Explain plans flag sequential scans and expensive nodes
Lock waits show which queries were blocked, and on what

Query detailsPostgreSQLorders-db

SELECT o.id, o.total FROM orders o JOIN customers c ON c.id = o.customer_id WHERE c.email = $1Requests312Avg latency18.40 msTotal time5.7k ms% time8.9%

Sampled executions by wait groupIO 62%CPU 31%Lock 7%

Execution plan

Nested Loopcost 4812.31└─Seq Scanon customers ccost 4793.10Seq scan└─Index Scanusing orders_customer_id_idx on orders ocost 8.44

Connect

One form per database, no agent on the server

Connecting a database adds it to a collector policy. The collector signs in as the monitoring user you create, and the queries it runs for monitoring stay out of your query views.

PostgreSQL
Top queries and plans need pg_stat_statements and the pg_monitor role; samples work without them.
MySQL
Query digests, samples and waits come from Performance Schema.
SQL Server
Live sessions and waits, and top queries from Query Store and the execution statistics.
Redis, MongoDB, Elasticsearch
Health metrics for memory, connections, replication, operations and more.

orders-dbPostgreSQL 16 · pg_stat_statements on

monitoring user · every 30 seconds

Collector web-prod-14Policy databases, on a network that reaches it

metrics, top queries, samples and plans

acme-prod workspaceLoad, queries, plans and recommendations

Usage from traces

See which services call each database

Instrumented services already send a span for every database call. Tracix reads them with the OpenTelemetry database conventions, for any engine, and shows throughput, errors and latency by database, operation and calling service.

Works for databases you have not connected, as long as your services are traced
Each database instance lists the services that call it, with their statements
Ready-made monitors you can turn on cover deadlocks, replica lag, connections and cache hit rate

Usage from tracesDatabase client spans from instrumented services

Databases4Queries412kThroughput458/sError rate0.02%Worst p9941 ms

orders-dbPostgreSQLCalling services

Service and operationCallsAvgp95

checkout-api43.6k0.6 ms2.9 ms

SELECT carts41.2k0.4 ms

UPDATE orders2.1k2.9 ms

SELECT orders31218.4 ms

cart-api19.6k0.9 ms2.2 ms

SELECT promotions19.6k0.9 ms

Better together

A slow query shows up everywhere at once

Database telemetry shares a workspace with traces, metrics and monitors, so you can follow a slow page to the query behind it and keep watching it afterwards.

Supported engines

What Tracix reads from each database

Health metrics come from the OpenTelemetry receiver for each engine. Query-level monitoring covers the three relational engines.

Engine	Health metrics	Top queries and plans	Samples and wait events	Replication lag	Schema objects
PostgreSQL	Yes	Yes	Yes	Yes	Tables, indexes, functions
MySQL	Yes	Yes	Yes	No	Tables
SQL Server	Yes	Yes	Yes	No	No
Redis	Yes	No	No	No	No
MongoDB	Yes	No	No	No	No
Elasticsearch	Yes	No	No	No	No

Health metrics are collected every 30 seconds by default. Usage from traces works for any database your instrumented services call, whether or not it is connected.

FAQ

Questions database teams ask first

Talk to an engineer

Which databases does Tracix support?

Health metrics for PostgreSQL, MySQL, SQL Server, Redis, MongoDB and Elasticsearch. Query-level monitoring, with top queries, samples, wait events and explain plans, covers PostgreSQL, MySQL and SQL Server.

Do I install anything on the database server?

No. A collector that can reach the database connects over the network as a monitoring user, with TLS and client certificates if you turn them on. The queries it runs for monitoring are left out of your query views.

What does PostgreSQL need for query monitoring?

The pg_stat_statements extension and the pg_monitor role for top queries and explain plans. Live query samples and wait events work without them.

Are the recommendations written by AI?

No. They come from rules over your query telemetry, such as expensive scans, missing or unused indexes, regressions, lock contention and long-running queries. When an AI model is configured, Brain adds a short written summary.

Connect your first database in minutes

Create a monitoring user, pick the collector policy that can reach the database, and the databases page fills in as the first collection runs.

Start free See the PostgreSQL integration

Know which query is slowing your database

From a monitoring user to the query that matters

Every normalized query, with its waits and its plan

One form per database, no agent on the server

PostgreSQL

MySQL

SQL Server

Redis, MongoDB, Elasticsearch

See which services call each database

A slow query shows up everywhere at once

What Tracix reads from each database

Questions database teams ask first

Connect your first database in minutes