Stable VBucket placement
A bounded virtual bucket space maps workflow_id → bucket_id → shard_id to a named execution boundary. Exact bucket counts and physical topology remain deployment- and version-bound implementation details.
MEMM exposes an implemented workflow/sharding foundation with stable VBucket placement, shard-local execution, explicit ownership and recoverable migration. The claim is deliberately limited to this engineering layer; it does not turn the runtime into a complete TI-Flow, HCC, ZETA or FHIR implementation.
Draft_TI-Flow_26_2 is used to classify surrounding service, HCC, ZETA and FHIR contexts. It is not treated as an approval or conformance statement.These statements apply to the workflow/sharding foundation itself and remain separate from higher-level TI-Flow product and provider responsibilities.
A bounded virtual bucket space maps workflow_id → bucket_id → shard_id to a named execution boundary. Exact bucket counts and physical topology remain deployment- and version-bound implementation details.
Workflow state, jobs, timers, subscriptions, journal and effect/outbox records stay on the responsible shard in the execution path; distributed JOINs and writes do not belong in the hot path.
Workers operate only inside their assigned shard-local execution boundary, where work is claimed and executed under explicit ownership. Wakeup, recovery and scheduling mechanisms remain internal runtime details.
Ownership, leases and fencing keep execution on the named shard and expose shard/worker failure as an explicit operating state.
Bucket migration follows an explicit recoverable sequence of draining, copy, verification and locked cutover with restart/recovery semantics.
VBuckets are routing and placement primitives. Bucket ranges do not automatically become medical, accounting, privacy or tenant boundaries.
Migration is a controlled protocol. Sensitive phases may deliberately fail fast or return retryable outcomes to avoid split-brain or ambiguous ownership.
Checkpointed phases allow a coordinator to determine a clear continuation or error state after restart.
Before verified cutover, source ownership remains authoritative; network or database failure cannot be interpreted as successful migration.
Versioned topology, locked cutover and fencing are used to prevent simultaneous active owners for one bucket domain.
The 30 June 2026 gematik pre-release covers service concepts, HCC clusters, onboarding, product/provider profiles and multiple FHIR Implementation Guides. Those layers are not automatically derived from a BPMN/VBucket runtime.
Unknown shards, missing owners, stale ownership and unresolved migration remain explicit states. ZETA-specific validation is not inferred from this runtime behavior.
Incomplete or contradictory active topology is not treated as normal routing.
Local/remote ownership and fencing restrict execution to the named owner.
Auth, tenant/project and request metadata stay inside the routing boundary.
ZETA remains external TI 2.0/TI-Flow context; complete ZETA validation or conformance is not claimed.
The view is synthetic. It does not show a real TI-Flow instance, gematik review or measured operating data.
Illustrative perspective on process state, technical execution and operating boundaries.
Separate FHIR/service contract, security/HCC context, workflow translation and the implemented runtime foundation against your target environment.