Performance sampling
The Performance tab collects metrics from the connected database and renders live charts. No agent or background process is required — sampling runs while you are connected.
Sample metrics
Each sample captures:
| Metric | Source |
|---|---|
| TPS (transactions per second) | Derived from transaction counters |
| Avg query response time | pg_stat_statements mean execution time |
| Active / idle / idle-in-tx connections | pg_stat_activity |
| Cache hit ratio | Derived from PostgreSQL block hit/read counters |
| WAL LSN | pg_wal_lsn_diff |
| Database size | pg_database_size |
| Checkpoints (requested + timed) | pg_stat_bgwriter |
| Temp files and bytes | pg_stat_database |
Storage
- Local browser storage — recent samples are retained locally so charts survive tab refreshes.
- Saved instance history — for saved instances, pocketPG can sync recent history server-side so reconnects have more context.
Charts combine recent local and current-session samples for the full picture.
Windowed analysis
Health and Tuning use recent performance samples to avoid making decisions from stale workload behavior.
The app requires enough samples before calculating rates such as checkpoint frequency, temporary file growth, and WAL generation.
Disk growth forecast
The disk forecast estimates recent growth trend from database size samples. It is meant as a planning hint, not a storage-capacity guarantee.
I/O write pressure
The I/O write pressure card highlights when write activity appears high enough to affect latency or checkpoint behavior.
Commit ratio
The commit ratio card suppresses alerts on idle servers so low-volume activity does not create false alarms.