Two-node keep-oldest could NEVER survive a crash of the oldest/active node:
Akka.NET 1.5.62 KeepOldest.OldestDecision only lets down-if-alone rescue a
side with >= 2 members, so the 1-vs-1 survivor takes DownReachable and downs
ITSELF — proven live on the rig ('SBR took decision ... including myself')
before this change. static-quorum(1) is worse (IsTooManyMembers -> DownAll);
keep-majority just re-keys the fatal crash to the lowest address.
SplitBrainResolverStrategy gains 'auto-down' (new default): BuildHocon emits
Akka's AutoDowning provider with auto-down-unreachable-after = StableAfter.
The leader among the REACHABLE members downs the unreachable peer, so the
survivor takes over singletons and /health/active in ~25s regardless of which
node died. Accepted trade (explicit owner decision): a real network partition
runs dual-active until an operator restarts one side. keep-oldest remains
supported; DownIfAlone validation is now scoped to it.
Live drill on the rebuilt rig: active-crash TAKEOVER in 28s (victim still
down; all 7 singletons Younger->Oldest), standby-crash removal 27s with 0
routing blips; victims rejoin as standby in 2s. New real-cluster tests pin
both directions (SbrFailoverTests.AutoDown_*); TwoNodeClusterFixture gains a
strategy knob. All 16 appsettings flipped (src, docker, docker-env2, and the
gitignored wonder-app-vd03 overlay on disk — owner must sync to the host).
Docs: decision record docs/plans/2026-07-21-auto-down-availability-decision.md,
Component-ClusterInfrastructure downing section rewritten, drill + README
reworked (active mode now asserts takeover), deferred-work SBR row resolved.
WhenTerminated watchdog calls IHostApplicationLifetime.StopApplication() when
the ActorSystem dies outside StopAsync (SBR self-down + run-coordinated-shutdown-when-down),
so the service supervisor restarts the node as a fresh incarnation. Optional
ctor param keeps every existing construction site compiling.
Plan's deploy/wonder-app-vd03/install.ps1 service-recovery-actions edit is
skipped: that production deploy artifact is not tracked in this repo (Task 16
skipped its appsettings.Central.json for the same reason).
Final cross-bundle reviewer identified 7 inconsistencies that the per-bundle
reviewers couldn't see; all fixed in one logical commit.
Critical:
- HighLevelReqs AL-3: drop 'then upsert-on-newer-status' — AuditLog is
strictly append-only (correct for SiteCalls/Notifications, wrong for
the immutable AuditLog shadow).
- Component-AuditLog Error rate KPI: align with HealthMonitoring's
exclusion list (Success/Delivered/Enqueued) rather than just non-Success;
otherwise every Delivered notification or Enqueued cached call would be
counted as an error.
Important:
- Component-AuditLog line 154: ISiteAuditWriter -> IAuditWriter (canonical
name per Commons and the rest of this doc).
- Component-AuditLog Central direct-write paragraph: convert remaining
slash notation (ApiInbound/Completed, Notification/Attempt,
Notification/Terminal) to dot notation used everywhere else.
- Component-ClusterInfrastructure: scope SiteCallAuditActor to
reconciliation + KPIs + Retry/Discard relay; cached-telemetry ingest is
AuditLogIngestActor's role per Combined Telemetry contract.
- Component-CentralUI Audit Log page: state the OperationalAudit read
permission and the read-vs-export split (matching CLI doc).
- Component-NotificationOutbox: add never-fail-the-action invariant for
dispatcher audit writes.
Minor:
- Component-InboundAPI: 'Non-blocking semantics' was ambiguous (could be
read as async); reword to 'Fail-soft' — the write is still synchronous
before flush, but failures are caught and don't change the response.
- Component-CLI: realign audit-query/audit-export flags to actually match
the Central UI Audit Log filter set (channel, kind, status, site,
instance, target, actor, correlation-id, errors-only); drop --user and
--entity-id which are IAuditService concepts, not Audit Log columns.
- Component-AuditLog KPI tile names: 'Volume/Error rate/Backlog' ->
'Audit volume/Audit error rate/Audit backlog' (matches Central UI and
Health Monitoring); drop the two orphan KPIs (Top inbound callers, Top
outbound 5xx) that were never surfaced anywhere.
- Component-AuditLog Interactions: re-attribute DbOutbound emissions to
ESG (where Database.* lives) with a note that Site Runtime is the API
surface for scripts.
- HighLevelReqs AL-12: drop 'and reconciliation operations' (CLI has no
reconcile command; reconciliation is an internal self-healing pull).
Add note that verify-chain becomes operational once AL-11's hash chain
ships.
Reviewer flag on 1bbfad3: "per Component-AuditLog.md, §6.2" pointed at
alog.md numbering, not at any anchor in Component-AuditLog.md (which uses
prose subsection titles). Switch to the prose anchor (Ingestion Paths →
Telemetry forward) so the link resolves.
LmxProxy is no longer needed. Moved the entire lmxproxy/ workspace, DCL
adapter files, and related docs to deprecated/. Removed LmxProxy registration
from DataConnectionFactory, project reference from DCL, protocol option from
UI, and cleaned up all requirement docs.
Switch site host to WebApplicationBuilder with Kestrel HTTP/2 gRPC server,
add GrpcPort/keepalive config, wire SiteStreamManager as ISiteStreamSubscriber,
expose gRPC ports in docker-compose, add site seed script, update all 10
requirement docs + CLAUDE.md + README.md for the new dual-transport architecture.
Organize documentation by moving requirements (HighLevelReqs, Component-*,
lmxproxy_protocol) to docs/requirements/ and test infrastructure docs to
docs/test_infra/. Updates all cross-references in README, CLAUDE.md,
infra/README, component docs, and 23 plan files.
Joseph Doherty