Recipe: Multi-instancia con 1 collector¶

Monitor N PostgreSQL instances with a single collector process (BON-013). Every metric, event and span carries a db_instance label.

1. Configure instances¶

Replace BONITO_DSN with BONITO_INSTANCES (JSON):

export BONITO_INSTANCES='[
  {"name": "app",  "engine": "postgresql", "dsn": "postgresql://bonito_ro:...@db1:5432/app"},
  {"name": "app2", "engine": "postgresql", "dsn": "postgresql://bonito_ro:...@db2:5432/app2"}
]'
bonito-collector

BONITO_DSN still works for the single-instance case (instance name default).

2. What changes¶

Output	Effect
Prometheus gauges	label `db_instance="app"` / `"app2"`
JSON events	`db_instance` field on every event
OTLP spans	`db.instance` attribute
API	`db_instance` query param (default `default`)

3. Query per instance¶

curl -H "X-API-Key: $BONITO_API_KEY" "localhost:8100/queries/top?db_instance=app&limit=5"
curl -H "X-API-Key: $BONITO_API_KEY" "localhost:8100/locks/active?db_instance=app2"

4. Dashboards & alerts¶

Grafana panels group by db_instance ("Active sessions by instance"), and alert rules carry db_instance in labels so Remo enrichment queries the right instance:

bonito_blocking_sessions{db_instance="app"} 1.0   # alert labels include db_instance=app

Verified¶

The demo tenant runs two instances (app, app2) — 10 metric series each in Prometheus, both visible in the "Deep DB View" dashboard.