Read replicas: what each plan gets (roadmap)
Read replicas exist today as connection strings; this is the plan for routing data-API reads to them, read-your-writes, lag limits and cross-region replicas, per plan.
Today a dedicated database (Pro and up) can add streaming read replicas and connect to them with their own connection string. The data API, realtime and functions still read from the primary. Planned next:
| Free | Pro | Ultra | Team | Enterprise | |
|---|---|---|---|---|---|
| Replicas per project (live today) | 0 | 1 | 2 | 5 | 10 |
| Data-API read routing | — | when asked | automatic | automatic | automatic |
| Max lag before a replica is skipped | — | 5 s | 2 s | 1 s | 1 s |
| Replicas in other regions | — | — | — | 2 | 5 |
| Failover to a replica | — | — | — | yes | yes |
- Only single read-only statements outside transactions go to a replica; anything else, and anything that fails on the replica, runs on the primary.
- Read-your-writes: after a write the SDK sends its write position for a few seconds; a replica that hasn't replayed that far is skipped.
- Replicas are billed by the hour like any database size.