39 lines
1.4 KiB
Markdown
39 lines
1.4 KiB
Markdown
|
# Changelog — BUG Batch 3：重试分类 / 推送计数 / 仪表盘在线（2026-06-04）
|
|||
|
|
|
|||
|
|
## 摘要
|
|||
|
|
|
|||
|
|
修复推送重试任务在「仅取消」场景下无限 pending、批量推送异常时成功/失败计数偏少、以及 Dashboard `online` 可能大于 `total` 的三处 P2 逻辑问题。
|
|||
|
|
|
|||
|
|
## 动机
|
|||
|
|
|
|||
|
|
- `retry_runner` 成功条件过窄，cancel-only 的 `sync_files` 结果永不结案。
|
|||
|
|
- `sync_engine_v2` 使用 `gather(return_exceptions=True)` 但未对缺失 `results` 的服务器补计失败。
|
|||
|
|
- Redis 残留 `heartbeat:*` 在服务器已删后仍计数，导致在线率超过 100%。
|
|||
|
|
|
|||
|
|
## 变更
|
|||
|
|
|
|||
|
|
| 文件 | 说明 |
|
|||
|
|
|------|------|
|
|||
|
|
| `server/background/retry_runner.py` | `apply_push_retry_outcome()`、`_backoff_seconds()` |
|
|||
|
|
| `server/application/services/sync_engine_v2.py` | gather 异常日志 + `unaccounted` 并入 `failed` |
|
|||
|
|
| `server/api/servers.py` | `online` 封顶为 `total` |
|
|||
|
|
| `tests/test_retry_runner_outcome.py` | 重试结果分类单测 |
|
|||
|
|
| `docs/reports/audit-bug-line-walk-2026-06-04-batch3.md` | BUG Batch 3 报告 |
|
|||
|
|
|
|||
|
|
## 迁移 / 重启
|
|||
|
|
|
|||
|
|
- 需重启 `nexus` Supervisor 进程。
|
|||
|
|
- 无 schema 变更。
|
|||
|
|
|
|||
|
|
## 验证
|
|||
|
|
|
|||
|
|
```bash
|
|||
|
|
ruff check server/background/retry_runner.py server/application/services/sync_engine_v2.py server/api/servers.py
|
|||
|
|
pytest tests/test_retry_runner_outcome.py tests/ -q
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
## 关联
|
|||
|
|
|
|||
|
|
- Batch 2：`docs/changelog/2026-06-04-schedule-once-rollback.md`
|
|||
|
|
- Batch 1：`docs/changelog/2026-06-04-install-lock-atomic.md`
|