Cold queries
Brussle keeps the namespaces that you query in a cache. A namespace that gets no queries for some time falls out of the cache. The next query must load the namespace first. We call this a cold query. A cold query takes some hundreds of milliseconds. See the latency targets.Pin a namespace
Setpinned to true in the namespace’s settings:
ns.update({ pinned: true }) (TypeScript) or ns.update(pinned=True) (Python). The response is the namespace with its new settings. GET /namespaces/{ns} and ns.metadata() also show pinned.
In the dashboard, open the namespace and use the Pinned switch at the top of the page. Owners and admins can change it.
What you can rely on
- After Brussle loads a pinned namespace, the namespace stays in the cache. Idle time does not remove it.
- The first query after you pin the namespace can still be cold. Pinning does not load the namespace. Warm the namespace after you pin it.
- A query after Brussle restarts a node can be cold. Warm the namespace again if you can.
- Pinning changes no answer and no document. It changes only how fast queries are after idle time.
Unpin a namespace
Setpinned to false. The namespace then leaves the cache when Brussle needs the room, like any other namespace. Brussle stops to bill the namespace as pinned at its next hourly measurement.
Price
Brussle bills a pinned namespace at $0.50 per GiB-month, on thepinned_bytes_hours line, on top of its storage. Brussle measures the namespace’s size each hour. See pricing.
For example, a pinned namespace of 2 GiB costs $1.00 a month to pin, plus its storage.