Database metrics

PandaStack exposes two distinct views of a running database: engine metrics (connections, queries per second, cache hit ratio) sourced from Managed Prometheus, and a live pod snapshot (phase, CPU, memory, restarts) from the Kubernetes cluster. Both are read live through worker-v3.

Engine metrics (paid)

Database engine metrics are a paid feature, sourced from KubeBlocks Managed Prometheus via the postgres / mysqld exporters. They describe how the database engine itself is behaving.

Summary

GET /v1/databases/:id/engine-metrics

Returns DatabaseEngineMetricsSummary:

Field	Type	Notes
`range`	enum	The time range (`1h`, `24h`, `7d`, `30d`)
`enabled`	boolean	Whether metrics are available/enabled
`engine`	string	`postgresql` or `mysql`
`up`	boolean (nullable)	Whether the engine exporter reports the DB up
`connections`	number (nullable)	Current active connections
`maxConnections`	number (nullable)	Configured connection ceiling
`connectionsPct`	number (nullable)	Connections as a percentage (0–100) of max
`qps`	number (nullable)	Queries / transactions per second
`cacheHitPct`	number (nullable)	Cache hit ratio, 0–100

All numeric fields are nullable — the UI shows - when no sample has been reported.

{
  "success": true,
  "data": {
    "range": "24h",
    "enabled": true,
    "engine": "postgresql",
    "up": true,
    "connections": 12,
    "maxConnections": 300,
    "connectionsPct": 4,
    "qps": 87.4,
    "cacheHitPct": 99.1
  }
}

Time series

GET /v1/databases/:id/engine-metrics/timeseries

Returns DatabaseEngineMetricsTimeseries — a series of points for charting connections, QPS, and cache hit ratio over the window:

Field	Type	Notes
`range`	enum	`1h` / `24h` / `7d` / `30d`
`enabled`	boolean	Whether metrics are available
`engine`	string	`postgresql` or `mysql`
`points`	array	See point shape below

Each DatabaseEngineMetricsPoint:

Field	Type	Notes
`t`	string	ISO bucket timestamp
`connections`	number (nullable)	Connections at that bucket
`qps`	number (nullable)	QPS at that bucket
`cacheHitPct`	number (nullable)	Cache hit ratio at that bucket

A null point means no sample was taken in that bucket, so the chart draws a gap rather than a fabricated value.

Ranges

Range	Window
`1h`	Last hour
`24h`	Last 24 hours
`7d`	Last 7 days
`30d`	Last 30 days

Live pod snapshot

Separate from engine metrics, PandaStack surfaces a live pod snapshot of the database’s Kubernetes workload, sourced through worker-v3 (/internal/db/metrics). This is DatabaseMetrics:

Field	Type	Notes
`available`	boolean	Whether a live snapshot was obtained
`phase`	string (nullable)	Pod phase, e.g. `Running`, `Pending`
`ready`	boolean	Whether the pod is ready
`restarts`	number	Container restart count
`startedAt`	string (nullable)	When the pod started
`cpuMilli`	number (nullable)	Current total CPU in millicores
`memoryMiB`	number (nullable)	Current total memory in MiB
`containers`	array	Per-container CPU/memory

Each DatabaseContainerMetric: name, cpuMilli, memoryMiB.

{
  "success": true,
  "data": {
    "available": true,
    "phase": "Running",
    "ready": true,
    "restarts": 0,
    "startedAt": "2026-07-14T02:00:00.000Z",
    "cpuMilli": 45,
    "memoryMiB": 312,
    "containers": [
      { "name": "postgresql", "cpuMilli": 40, "memoryMiB": 300 },
      { "name": "metrics-exporter", "cpuMilli": 5, "memoryMiB": 12 }
    ]
  }
}

Databases overview
Backups & restore — also read live via worker-v3.
Tiers & scaling — connection limits by plan.
Plans & limits — analytics/metrics as paid features.