Scale to zero, keep-warm and idle pause
NamiDB Cloud runs your organization’s engine only while it is used. Two separate mechanisms are involved, and they are easy to confuse:
| Scale to zero | Idle pause | |
|---|---|---|
| Plans | All | Free only |
| After | 10 / 30 / 60 idle minutes (Free / Pro / Scale) | 7 days without any activity |
| What happens | The engine stops; your data stays where it is | The organization is paused; queries are refused |
| Next query | Wakes it automatically | Answers 403 org_paused until someone resumes it |
| Billed? | No compute charge either way: you pay for storage and requests | Nothing is served, so nothing is billed |
Scale to zero
When no request reaches any database of your organization for the plan’s idle time, its engine stops. The next request starts it again, transparently:
- HTTP and MCP requests wait while the engine starts, then run. If the
start takes longer than a request may wait, the request answers
503 engine_startingwithRetry-After: 1. The SDKs retry it automatically, writes included (nothing ran). - Bolt connections wait while the engine starts, then carry on as usual.
- The console’s Activity and Metrics tabs never wake the engine on their own.
Nothing is lost when an engine stops: every acknowledged write is already in your organization’s storage.
Keep warm (Scale)
Scale organizations can turn on Settings → Engine → Keep warm (owners and admins). The engine is then never stopped for being idle, so there are no cold starts, and it is started again on its own if it ever stops. The setting has no effect on other plans; moving away from Scale turns it off in effect.
Designing for cold starts
- Set client timeouts above the plan’s query timeout plus a few seconds. The SDKs default to 90 seconds per attempt.
- Treat
engine_startingandrate_limitedas retryable (the SDKs do). - For latency-sensitive production traffic on Pro, send a cheap query
(
RETURN 1) on a schedule shorter than 30 minutes, or move to Scale with keep-warm.
Idle pause (Free)
A Free organization is paused after 7 days without activity. Activity is any of: an API key used, a request served, or a member signing in to the console. Creating or resuming the organization also counts.
- Owners get an email 24 hours before the pause, and another when it happens.
- While paused, every query answers
403 org_paused(over Bolt, aSecurity.Forbiddenfailure and the connection closes). The data and backups stay untouched. - Resume from the banner in the console (owners and admins); queries work again within seconds. Upgrading to a paid plan also resumes it.
- A Free organization paused for 90 days may be deleted, after 30 days’ notice by email (sent at 60 days paused, with a reminder a week before). Nothing is deleted without that notice.
Paid plans are never paused for inactivity.