Chaos Engineering Platform  ·  View 10 of 21  ·  Data

Data Architecture — Ownership and Rebuildability

Four zones, graded by who wrote the data and what losing it costs.

Editable source SVG draw.io All views
Authored elsewhere — read only Consumed inventory Service catalogue 24 h staleness cap Dependency graph from mesh traffic SLO definitions the owners' SLIs Owned by the platform Hot, strongly consistent — not rebuildable Active leases Spanner, regional Radius reservations Spanner Chaos budget ledger Spanner Durable record — RPO 0, append-only Run records 24 months Findings to closure Audit log 7 years Derived — rebuildable from the record Coverage marts BigQuery Cost attribution BigQuery Sampled — loss degrades a verdict, never corrupts one High-res signals 90 days, RPO 60 s Downsampled series life of the run Sourced from the definition repo Version-controlled Definitions immutable per version Templates version pinned after 90 days rebuildable terminal state Data Architecture — Ownership and Rebuildability External / third party Data store synchronous batch Losing the derived zone costs a rebuild. Losing the lease zone is safe: every fault reverts. v 1.0 · owner Reliability Architecture · date 2026-09

Decisions

  • Losing the lease zone is safe by construction: every fault reverts. That is why the hottest, most consistency-sensitive store in the design is also the one with the cheapest failure mode.
  • Run records, findings and audit entries commit synchronously — RPO 0 — because they are the only durable claim that a resilience test happened.
  • Sampled signals are explicitly second-class: their loss changes a verdict rather than corrupting one.

Assumptions

  • Run records and verdicts 24 months; audit 7 years; high-resolution signals 90 days at RPO 60 s.
  • Coverage and cost marts rebuildable from the run record within one working day.

Deliberate omission

  • Backup topology and export formats are not drawn. They follow the standard platform pattern for Spanner and BigQuery and would add nothing to this argument.
◀  Data Flow — Definition to Evidence All views Data Model  ▶
Solution Architecture v1.0 · Google Cloud · Reliability Architecture · 2026-09 · 21 views · 16 architecture decision records10-data-ownership-zones