Overspan.dev

DOC Exit path

Migrating away from Overspan

We keep this page current because lock-in by inconvenience is not a product. If Overspan stops being the right answer for you, here is how to leave cleanly.

The short version

There is nothing to export. Overspan stores no data of yours: not your queries, not your results, not a database in some proprietary shape. Your queries are standard Overpass QL, the same text you would send anywhere else. Migration is changing one URL in your configuration.

the entire migration
- https://api.overspan.dev/YOUR_KEY/api/interpreter
+ https://overpass-api.de/api/interpreter

The one thing to check is your [timeout:] and [maxsize:] settings, since the destination will have its own limits. See the notes below.

Cancelling

Cancel yourself in the billing portal. Enter the email you bought with and it sends you a sign-in link, so there is no password to dig out. No notice period, no retention call, no form to fill in.

  • Your key keeps working until the end of the period you have paid for. Cancelling on day two of a month does not cut you off on day two.
  • When that period ends, the key stops. Requests then return 401 invalid_key.
  • Changed your mind before the period ends? Resume in the same portal and nothing was interrupted.
  • Invoices and receipts are in that portal too, along with your payment method.
  • Want your usage history? Ask us before you cancel and we will send you your own request records. After the subscription ends we do not keep them indefinitely; see the privacy page.

Back to the public servers

Entirely reasonable, and for plenty of workloads it is the right call. Point your client back at https://overpass-api.de/api/ or another instance from the OSM wiki's list, and undo whatever base-URL change you made when you arrived.

What changes when you go back:

  • Timeouts differ. The public servers default to 180 seconds where we inject 25, so queries that relied on our default will behave differently. Setting [timeout:] explicitly makes your queries portable everywhere, which is worth doing regardless.
  • You share a queue again. Expect 429 and 504 under load, and expect the limits to move without notice.
  • Cloud IPs may be blocked. If you moved to us because your CI or Lambda was refused, that will come back.
Be a good neighbour on the way back The public instances are donated infrastructure. Rate-limit yourself, cache aggressively, avoid running production traffic through them, and consider donating to OSM. That commons is upstream of this entire product, ours included.

To another provider

Other commercial and community Overpass instances exist, with different pricing shapes: community-gated hobby instances, and enterprise providers selling annual contracts with volume commitments. Any of them that runs the standard Overpass software takes the same QL you are already sending.

Worth checking before you commit, in roughly this order:

  • Metadata. Some instances run with meta=no, which silently drops version, timestamp, changeset, user, and uid from results. If you use any of those, confirm before switching.
  • Freshness. Ask how far behind the replication feed they run. Nightly is common and is very different from minutely.
  • Areas. Area queries need generated area files. Not every instance builds them.
  • Their caps. Compare their per-query timeout and memory ceilings against the heaviest query you actually run, not your average one.

Running your own

Completely viable, and if your volume is high enough it is cheaper than paying anyone. This is the same software we run, so nothing about your queries changes.

What you are taking on, honestly:

  • A full-planet clone with metadata lands somewhere near half a terabyte on disk and grows every year.
  • Enough RAM that queries are not disk-bound, which is the difference between fast and unusable.
  • A minutely update loop that must be monitored, because the standard failure mode is that it silently stops while the server keeps answering every query perfectly from frozen data.
  • Roughly a day of initial sync before it serves anything.

The Overpass installation guide and the community Docker images are the place to start. If you go this way and get stuck on something we have already hit, mail us. We would rather you succeed than churn back unhappy.

Checklist

  1. Set [timeout:] and [maxsize:] explicitly in every query, so behaviour does not depend on any server's defaults.
  2. Point a staging environment at the new endpoint and compare results on a query you know well.
  3. Confirm metadata fields still appear if you depend on them.
  4. Move production traffic across.
  5. Request your usage history if you want it.
  6. Cancel in the billing portal. Your key runs to the end of the paid period.
  7. Remove the key from your secret store and any environment configuration.

Leaving because something broke or because a limit was wrong for you? Mail hello@overspan.dev and tell us which. We would rather hear it than guess.