Storage and Durability | TraceDB Documentation

Local or mounted storage remains a first-class runtime backend. Hosted TraceDB Cloud uses AWS S3 for durable object storage and Railway only for stateless compute and cache/spool.

v0.1.1 may document the public runtime storage contract, AWS/IAM baseline, S3 object-operation expectations, manifest compare-and-swap, and lease-fencing model. Full hosted runtime rehydrate from S3-backed WAL/checkpoint state after an engine restart is v0.1.2 work, not a v0.1.1 release blocker.

Local WAL, checkpoints, manifests, snapshot/restore, and idempotency receipts remain runtime-owned behavior. Hosted provider wiring must not be described as a managed DR/SLA guarantee until separate validation and operator receipts exist.