System — Host-Level Metrics (CPU, RAM, Disk)
Cloud provider host metrics + buffer cache breakdown + filesystem usage.
What you see on this page depends on which cloud integration you've set up. The PostgreSQL-internal metrics are always there; the host-level metrics (CPU, memory, IOPS) come from your cloud's monitoring API or a node exporter for self-hosted.
| Provider | Source | Metrics |
|---|---|---|
| AWS RDS | CloudWatch (needs IAM) | CPUUtilization, FreeableMemory, ReadIOPS, WriteIOPS, NetworkRx/Tx, EBSIOBalance% |
| Azure Flex | Azure Monitor | cpu_percent, memory_percent, iops, storage_used, network_bytes_egress |
| Cloud SQL | Cloud Monitoring | cpu/utilization, memory/quota, disk/utilization, network/traffic |
| Self-hosted | node_exporter (if installed) | CPU per-core, RSS, load avg, disk usage, network counters |
Buffer cache (with pg_buffercache)
If you've installed pg_buffercache, the System page also shows what's actually in shared_buffers — top 20 relations by buffer count, dirty page count, utilization percentage. Without the extension, those cards stay empty. See setup.
The patterns to look for
CPU sustained over 80% with low TPS usually means sequential scans on large tables — check the Queries page for the offender. FreeableMemory dropping rapidly points to work_mem too high under load, or a memory leak in a function. EBSIOBalance dropping (RDS only) is gp2 burst credits running out — upgrade to gp3 or a higher IOPS class. Storage at 80%+ means autovacuum lag (bloat) or unbounded WAL accumulation; both have a happy ending if caught early.