Neon Postgres Dashboard - Monitor Connections, Cache, CPU

Use this dashboard to monitor your Neon projects across connections, database activity, the Local File Cache, connection pooling, read replica lag, and compute resources.

Neon sends these metrics through its built-in OpenTelemetry integration. You do not need to run an OpenTelemetry Collector.

Dashboard Coverage

Use these panels to:

• See database size, active connections, cache hit ratio, and CPU use.
• Track Postgres connections by state and by compute.
• Measure row changes per second, deadlocks, and database size.
• Make sure that the Local File Cache holds your working set. The Local File Cache is a cache on each compute that keeps pages from Neon storage.
• Watch the PgBouncer pool behind the pooled connection string: clients, server connections, wait time, throughput, and query duration.
• Track replication delay on read replicas.
• Compare CPU, load, memory, and swap on each compute.

Metrics Included

Overview

• Database Size: Total size of all databases (neon_db_total_size).
• Active Connections: Postgres connections that run a query (neon_connection_counts).
• Cache Hit Ratio: Share of page reads that the Local File Cache serves (neon_lfc_hits, neon_lfc_misses).
• CPU Utilization: Busy share of compute CPU time (host_cpu_seconds_total).

Connections

• Connections by State: Postgres connections grouped by pg_stat_activity state.
• Connections by Compute: Total Postgres connections on each compute.

Database Activity

• Row Operations: Rows inserted, updated, and deleted per second (neon_pg_stats_userdb).
• Deadlocks: Deadlocks per database in each 5-minute window.
• Size by Database and Total Database Size: Size of each tracked database and of all databases.

Neon collects neon_pg_stats_userdb only for its oldest non-system databases, up to a limit it does not publish. If your project has more databases than that limit, Row Operations, Deadlocks, and Size by Database leave out the newer ones, but Total Database Size includes all databases.

Local File Cache

• Cache Hit Ratio: Cache hit ratio over time.
• Cache Operations: Cache hits, misses, and writes per second.
• Cache Used vs Limit: Cache in use on each compute, compared with the cache size limit (neon_lfc_used_pages, neon_lfc_cache_size_limit).
• Working Set Size: Approximate size of the pages that queries touch in 5-minute, 15-minute, and 1-hour windows, compared with the cache size limit (neon_lfc_approximate_working_set_size_windows).

Connection Pooler (PgBouncer)

• Client Connections: Active and waiting clients on the pooled connection string.
• Server Connections: Active, idle, and used connections from PgBouncer to Postgres.
• Max Client Wait: Longest time that a client waits for a server connection.
• Pooled Throughput: Transactions and queries per second through PgBouncer.
• Average Query Duration: Mean time for each pooled query.

Read Replicas

• Replication Delay (Bytes) and Replication Delay (Time): Replication lag on each read replica (neon_replication_delay_bytes, neon_replication_delay_seconds). If the project has no read replicas, these panels show no data.

Compute

• CPU Utilization by Compute and Load Average (1m) by Compute: CPU use and run-queue length on each compute.
• Memory Utilization by Compute and Swap Used by Compute: Memory use and swap on each compute.

Dashboard Variables

Use these filter variables:

• project_id: Filter by Neon project.
• endpoint_id: Filter by compute. Each read replica has its own compute endpoint.

Neon sets service.name on metrics to fixed values, so the dashboard filters by project_id and endpoint_id instead.

Related Dashboards

• PostgreSQL
• PgBouncer
• CloudNativePG