OpenStreetMap/Overpass
Servizio di query su OpenStreetMap, usato abbondantemente dalla comunità italiana.
Istanza italiana
Spazio su Cloud VPS ottenuto in T426790 nel progetto osmit (Server#Altri_server_in_Wikimedia_Cloud).
Istanza in fase di setup, non farci ancora affidamento per applicazioni di produzione.
L'URL dell'istanza è https://osmit-overpass.wmcloud.org/api/interpreter , per usarla per fare query:
- Aprire https://overpass-turbo.eu/
- Cliccare su "Impostazioni"
- Nel campo "Server" usare il valore https://osmit-overpass.wmcloud.org/api/
- Cliccare salva
- Usare normalmente Overpass per eseguire le query
Link utili
- osm:Overpass
- osm:Overpass API
- Setting up an Overpass API server - how hard can it be?
- Repo originale drolbr/Overpass-API
- Fork aggiornato b1tw153/Overpass-API con istruzioni molto chiare e estensive
- Immagine Docker wiktorn/overpass-api
Requisiti
Informazioni dai requisiti ufficiali:
Storage
For a full planet import, expect roughly 4-5× expansion from the compressed `.osm.bz` file to the database. Check the current planet file size on planet.openstreetmap.org before provisioning storage.
Approximate full planet database sizes by metadata mode as of Q2 2026:
| Mode | Database Size |
|---|---|
| attic (full history) | ~750 GiB |
| meta (latest metadata) | ~365 GiB |
| no (base data only) | ~270 GiB |
| --areas=yes (area data) | ~30 GiB (additional) |
For extracts, the database will be proportionally smaller, but the expansion ratio from source to database is similar. Check the size of your chosen extract file and apply a 4-5× multiplier as a planning estimate. Note that most extract sources do not include full history, so `attic` mode is generally not available for extracts; `meta` mode is typical.
CPU
The container runs one fcgiwrap worker per CPU core plus one by default (`FCGIWRAP_WORKERS` defaults to `cpu.max`/`nproc` + 1). Each worker handles one concurrent query. Provision CPU cores based on your expected concurrent query load. A single core is sufficient for light personal use; a public-facing instance benefits from more cores to handle concurrent requests without queuing or rate limiting. An optional request queue can be configured in nginx.
RAM
- A minimal development instance with a small extract can run with 1-2 GiB
- A lightly loaded personal instance can run with 4-8 GiB
- A moderately loaded instance with complex queries benefits from 16-32 GiB
- A heavily loaded public instance may need 64 GiB or more
Set the container memory limit explicitly — for example, with `docker run --memory=16g`. Without a memory limit, the container can consume all available host memory under heavy query load.