Know Which Query is Killing Your Database - Every Second

Netdata provides native collectors for all major databases:

Full Top Queries support: PostgreSQL (pg_stat_statements), MySQL/MariaDB (performance_schema), SQL Server (Extended Events), Oracle (V$SQLSTATS), MongoDB (system.profile), Redis (SLOWLOG), ClickHouse (system.query_log), Elasticsearch (slow log), CockroachDB (crdb_internal), Couchbase, ProxySQL, and YugabyteDB.

Metrics and alerts: Cassandra, Memcached, RethinkDB, PgBouncer, MaxScale, Pika, and more.

Generic SQL collector: Works with any SQL-compliant database for custom metric queries.

All databases are auto-discovered with zero configuration required.

Top Queries is a built-in function that retrieves query performance data directly from your database’s internal statistics. For PostgreSQL, it queries pg_stat_statements. For MySQL, it uses performance_schema.events_statements_summary_by_digest. For SQL Server, it reads from Extended Events. For Oracle, V$SQLSTATS. For MongoDB, system.profile. You can sort by CPU time, elapsed time, logical reads, physical reads, executions, or rows returned to instantly identify the worst-performing queries.

Netdata tracks replication metrics every second. For PostgreSQL: streaming replication lag in bytes and seconds, WAL positions, replica state. For MySQL: seconds behind master, relay log position, Galera cluster status and flow control. For MongoDB: replica set member health, oplog position, replication lag. Alerts fire immediately when lag starts growing - not minutes later when it’s already a problem.

Yes. For SQL Server, Netdata retrieves deadlock graphs from the system_health Extended Events session. For PostgreSQL, it monitors pg_stat_activity for blocked queries and deadlocks from pg_stat_database. For MySQL, it tracks InnoDB deadlock counts via innodb_deadlocks. You get immediate visibility when deadlocks occur, not after users complain.

Netdata trains 18 machine learning models per metric. Each model learns what ’normal’ looks like for that specific metric - your query rates, your connection patterns, your buffer hit ratios. When behavior deviates from the learned pattern, it flags an anomaly. The 18-model consensus means a metric is only flagged when multiple models agree, achieving 99% false positive reduction compared to static threshold alerts.

For each database, Netdata collects comprehensive metrics:

PostgreSQL (85+): Connections, transactions, WAL, locks, replication lag, vacuum progress, cache hit ratios, table/index statistics, bloat, tablespace usage.

MySQL (100+): Connections, queries, threads, InnoDB buffer pool, handlers, binary log, Galera cluster, slave status.

SQL Server (50+): Connections, batch requests, buffer cache, page life expectancy, wait statistics, transaction log.

Oracle (70+): Sessions, tablespaces, SGA/PGA, cache hit ratios, wait classes, ASM.

MongoDB (60+): Operations, WiredTiger cache, connections, replication, sharding.

60 seconds. Run the one-line installer, and Netdata auto-discovers your databases and begins collecting metrics immediately. Dashboards are ready instantly. Pre-configured alerts activate automatically. ML models start training and produce anomaly detection results after 15 minutes of baseline data. No configuration files, no YAML, no query languages to learn.

Netdata uses read-only queries to collect metrics. For PostgreSQL, a user with pg_monitor role is sufficient. For MySQL, PROCESS and REPLICATION CLIENT privileges. For SQL Server, VIEW SERVER STATE. For Oracle, SELECT on system views. For MongoDB, clusterMonitor role. All connections can be read-only with least-privilege access.

Install Netdata on each database server, and it auto-discovers and monitors databases locally. Connect to Netdata Cloud for a unified dashboard showing all databases across all servers. Or deploy a Netdata Parent for centralized storage if you prefer to keep everything on-premises. Either way, you get a single view of your entire database estate.

Yes. Netdata exports to Prometheus, Graphite, OpenTSDB, InfluxDB, and TimescaleDB. It has a native Grafana datasource plugin. You can use Netdata’s per-second collection alongside your existing dashboards, or gradually consolidate as you see the value of unified database and infrastructure monitoring.

Every Netdata agent stores metrics locally on the database server itself. If the network goes down, if the cloud has issues, if parents are unreachable - collection continues. When connectivity returns, you have the complete picture of what happened during the outage. This is critical for root cause analysis after incidents.

Netdata provides three key advantages: (1) Per-second granularity catches issues that 30-60 second tools miss. (2) Unified monitoring - databases, servers, containers, and Kubernetes in one platform instead of separate tools. (3) Predictable pricing at 90% lower cost with no per-metric or per-query charges. The trade-off: Netdata doesn’t have some specialized features like historical query plan comparison that dedicated SQL Server tools offer.

Yes. Netdata agents work completely standalone with local dashboards at http://localhost:19999. For multi-server views, deploy Netdata Parents on-premises. Netdata Cloud On-Premises is available for enterprises that need the full control plane (RBAC, SSO, AI features) within their own datacenter. All observability data stays on your infrastructure.

Less than 5% CPU and 150-200MB RAM. Netdata uses efficient native code (C and Go), not Python. Database queries are lightweight and read-only. Independent research by the University of Amsterdam confirmed Netdata as the most energy-efficient monitoring solution. Monitoring doesn’t compete with your databases for resources.

Yes. Netdata includes 400+ pre-configured alerts, including many database-specific ones: connection pool utilization, replication lag thresholds, slow query rates, cache hit ratio drops, tablespace usage, deadlock occurrences, and more. You can customize thresholds or create new alerts. ML anomaly detection provides an additional signal by learning your specific patterns.