mirrors/seaweedfs
1
0
Fork 0
mirror of https://github.com/seaweedfs/seaweedfs.git synced 2026-08-21 22:56:55 +00:00
Code Issues Packages Projects Releases Wiki Activity

docs(recovery): record coordinator progress fact gaps

Browse Source
This commit is contained in:
pingqiu
2026-05-02 11:09:25 -07:00
parent 6905fb8278
commit baaa1e7e45
1 changed files with 15 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
sw-block/design/v3-phase-15-g7-progress-fact-recover-plan.md
+15
View File
@@ -8,6 +8,21 @@
---
## 0. Current State And Gap
The project already has coordinator layers; the issue is that their facts are still split.
| Layer | Current owner | What exists today | Gap |
|-------|---------------|-------------------|-----|
| Semantic coordinator | `core/engine` via `adapter.VolumeReplicaAdapter` | `ProbeResult` normalizes into `R/S/H`; engine decides `none / catch_up / rebuild / unknown`. | No unified progress fact type; decisions are tied to probe/session events directly. |
| Peer/probe host | `core/replication.ReplicationVolume` | Maintains peer set and degraded-peer probe loop; `G5_5C` covers restart → probe → auto catch-up. | Auto rebuild for no-trusted-R / R<S needs explicit component coverage. |
| Session/pin coordinator | `core/recovery.PeerShipCoordinator` | Owns active session phase, route-local-write, barrier witness, and pin floor. | Durable ack updates pin, but is not yet a first-class engine progress fact. |
| Feeder/executor | `core/transport` + `core/recovery` | Owns WAL/BASE movement and monotonic egress path. | Should remain execution-only; must not decide recovery class or membership. |
Immediate conclusion: **do not invent a new top-level coordinator**. Instead, add a shared progress fact seam so existing coordinators consume the same facts with explicit authority rules.
---
## 1. Target State
### 1.1 Roles