Monitoring and Metrics

Illustration by Annie Ruygt of a balloon doing a lot of tasks

Performance Monitoring

The Managed Postgres dashboard provides comprehensive performance monitoring and metrics for your PostgreSQL clusters. The Metrics tab gives you real-time visibility into your database’s performance, helping you identify bottlenecks and optimize your applications.

Accessing Metrics

To view metrics for your cluster:

Navigate to your MPG cluster in the Fly.io dashboard
Click the Metrics tab
Optionally filter by specific database using the database dropdown
Select your desired time range (5 minutes, 15 minutes, 1 hour, 6 hours, 24 hours, 2 days, or 7 days)

On older v1 clusters, the Metrics tab may show a New Metrics System Available card instead of charts. Enabling the new metrics system restarts all Postgres nodes in your cluster to apply the monitoring configuration. On v2 clusters, if a database machine was recently throttled significantly, a CPU throttling detected or Severe CPU throttling banner appears at the top of the tab.

Metrics tiles

The top of the Metrics tab shows four tiles with the cluster’s current state:

Tile	Description
Database Connections	Current connections to the selected database, against the cluster-wide `max_connections` setting
Pooled Connections	Current number of clients on the PGBouncer pool, active and waiting combined
Longest Query	Duration of the longest-running active query in the selected database
Replication Lag	Current maximum replay lag. The same measurement as the Replication Delay Seconds chart, as a single current value

Available Charts and Metrics

Below the tiles, each chart plots its series over the selected time range, with each series’ average shown underneath. The charts are grouped into these categories:

System Resource Metrics

Metric	Description
Database CPU Utilization	CPU usage percentage for each database instance, labeled with Primary (P) and Replica (R) badges. On v2 clusters this chart also shades periods when the database was CPU-throttled. A throttled machine can otherwise appear to sit at a flat 100% CPU
Database Memory Utilization	Memory usage over time for Primary and Replica database instances, labeled with instance IDs
Pooler CPU Utilization	CPU usage percentage for PGBouncer connection pooler instances. Typically remains low under normal operations
Pooler Memory Utilization	Memory consumption for connection pooler instances. Shows usage for each pooler with instance IDs

Connection Metrics

Metric	Description
Database Connections	Active and idle database connections over time
Pooler Connections	Tracks active and waiting connections through PGBouncer

Database Operations

Metric	Description
Database Operations	Row-level operations per second broken down into selects, inserts, updates, and deletes. Shows throughput patterns and workload distribution across operation types
Cache Hit Ratio	Proportion of block accesses served from PostgreSQL’s buffer cache. Reads outside that cache may still be served by the operating system’s page cache.
Deadlocks	Count of database deadlocks detected when two or more transactions are waiting for each other to release locks. Shows frequency and timing of these mutual blocking situations

Storage

Metric	Description
Database Size	Database size over time. Use the database dropdown to view a specific database

Replication Metrics

Metric	Description
Replication Delay Bytes	Amount of WAL (Write-Ahead Log) data in bytes that the replica is behind the primary. Measures the volume of changes waiting to be applied
Replication Delay Seconds	Time in seconds that the replica lags behind the primary database. Represents how long ago the replica’s current state reflects the primary

Overview tiles (v2 clusters)

The Overview page of a v2 cluster (marked v2 next to the cluster name) shows summary tiles with the cluster’s current state. v1 clusters do not have these tiles.

Tile	Description
CPU Usage	Current CPU use of the primary
Memory Usage	Current memory use of the primary
Storage Usage	Approximate storage consumed by user databases, shown against each node’s allocated storage. Excludes WAL and system overhead, so it does not represent total volume usage or remaining capacity
Active Connections	Open connections to your databases, including idle connections, against the cluster’s `max_connections` setting
Uptime	How long the cluster has continuously had a primary up, over the last 30 days

​Performance Monitoring

​Accessing Metrics

​Metrics tiles

​Available Charts and Metrics

​System Resource Metrics

​Connection Metrics

​Database Operations

​Storage

​Replication Metrics

​Overview tiles (v2 clusters)