Multi-region deployment
Cross-region logical replication, automatic failover, data residency by country. Patterns for global SaaS with local latency without sacrificing consistency.
Three common patterns#
- Single region (default) — everything in
fr-par(Scaleway, Paris): Postgres, Storage, Functions. Minimum intra-region latency, single point of failure at the region level. - Primary + read replicas — primary in one region, read replicas in other EU regions (fr-par → Germany, Finland). Centralized writes, distributed reads.
- Active-active multi-region (Enterprise) — Postgres with BDR (bi-directional replication), logical decoding per table, custom conflict resolution.
Table by table, filtered#
Enable logical replication from the Studio or CLI. Choose the tables to replicate (fine granularity: exclude heavy, non-critical tables).
Roadmap — not available. The aura commands shown below are planned and not delivered yet: the CLI does not have them today.
# Create a read-only replica in Germany
aura replica create \
--region de-fsn \
--tables orders,products,customers \
--mode logical
# Lag monitoring
aura replica status --region de-fsn
# → lag: 120ms · last_lsn: 7/3F2A12E8 · healthy
Automatic within 30 seconds#
Our agent monitors each primary with multi-AZ health checks. On confirmed failure (3 consecutive checks), a replica is promoted in ~30s. The gateway switches connections automatically.
During failover, writes are queued on the SDK client side (automatic retry with backoff). Readings continue on surviving replicas.
Pinning by country on Enterprise#
Free and Pro projects are hosted in the EU only. Plan Enterprise: pins each project to a country. Data, backups and replicas never leave the territory — contractual commitment, without third-party certification to date.
FR— France (GDPR)DE— Germany (GDPR)FI— Finland (GDPR)