root/fusion
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(FN-4044): harden task ID collision guards and allocation safety

Browse Source 
Hardens task ID collision guards across distributed instances by adding `isAllocated` checks in `store.ts` and rewriting `distributed-task-id.ts` with deterministic collision-safe allocation logic, backed by a new `store-create-collision.test.ts` suite reproducing the race path. Updates architecture

Fusion-Task-Id: FN-4044
This commit is contained in:
Fusion
2026-05-12 06:26:37 -07:00
committed by gsxdsm
parent e0fcc7e2fa
commit 72e94dddbe
11 changed files with 632 additions and 194 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
docs/storage.md
View File

@@ -4,8 +4,9 @@
- `distributed_task_id_state` is the authoritative local task-ID allocator state. `nextSequence` is the active high-water mark used for local ID reservations.
- `distributed_task_id_reservations` tracks reserve/commit/abort lifecycle entries. Aborted/expired reservations are burned and never reissued.
- `config.nextId` is retained only as a legacy compatibility field and optional seed source; runtime task creation no longer mutates it as allocator truth.
- Startup allocator reconciliation bumps each active prefix sequence to `max(current nextSequence, max(existing task suffix)+1)` across live + archived tasks to self-heal stale allocator drift.
- `config.nextId` is retained only as a deprecated legacy compatibility field and optional one-time seed source. Fusion still reads it during reconciliation, but runtime task creation and settings writes no longer mutate it.
- Startup/store-open allocator reconciliation bumps each active prefix sequence to `max(current nextSequence, max(tasks suffix)+1, max(archivedTasks suffix)+1, max(reservation sequence)+1)` so stale allocator rows self-heal before local task creation resumes.
- Create-class task persistence is intentionally non-destructive: new tasks use plain `INSERT` semantics, while `ON CONFLICT(id) DO UPDATE` remains update-only. If counters drift and a reserved ID still collides, the create fails and the existing SQLite row / task directory stays intact.
## SQLite write-path lock recovery (FN-4042 / FN-4083)