Kaotings/www_site
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
62dba45799
www_site/docs/OPERATIONS.md

87 lines
6.6 KiB
Markdown
Raw Normal View History

docs: Phase 5 acceptance, deployment, operations and phase report Record the task/quota exception matrix (isolated fault injection) results, the admin quota concurrency fix, the isolated restore drill (business/CMS/audio, with the pre-existing missing-audio finding), upstream security launch blocker, and remaining minor items. Each item marked pass / fail / awaiting-external with test version and evidence. Do not mark reliability/launch-blocking items complete.
2026-09-09 21:01:27 +00:00
# 运维方案（OPERATIONS）
日期：2026-09-10 · 版本 `8657f44`
范围：音频容量/保留/清理/备份/恢复、数据库备份/恢复、Worker 与额度对账、告警。**本轮不擅自删除任何既有音频。**
## 1. 音频存储现状
- 位置：`/home/flym/kaotings-audio`，布局 `{user_id}/{task_id}.wav`（私有本地目录，非对象存储）。
- 元数据：`audio_files`（`storage_key`、`mime_type`、`size_bytes`、`checksum`、`status`、`expires_at`）。
- 写入：成功结算时先写文件、再登记 `audio_files`（含 SHA256）。
- 读取：回放/下载按 `storage_key` 取文件；文件缺失返回 404（优雅降级）。
docs: record configured offsite NAS backup (rsync over SSH) Offsite backup to the Synology NAS (103.40.14.100:52200) is now configured: pg_dump of business+CMS DBs (keep 7 per series) plus rsync of the audio dir, encrypted over SSH with key auth, on a daily 03:00 systemd timer. Document the destination, transport, script, timer, and restore-from-NAS procedure; mark the backup-location item resolved (retention/threshold/owner still pending).
2026-09-10 07:24:04 +00:00
## 2. 备份（已配置异地备份）
- 本机暂存：`/home/flym/backups/`（仅存放 `.dump`）。
docs: correct offsite backup to /volume1/NetBackup via tar-over-SSH Moved the backup out of the NAS home dir to /volume1/NetBackup/kaotings per request. Synology restricts rsync-over-SSH to registered shared folders and scp/sftp yields root-owned files that the backup account cannot rotate, so the backup now uses tar over SSH (kts_bak-owned, rotatable, encrypted, key auth). Document the transport rationale and the DSM shared-folder upgrade path for incremental rsync.
2026-09-10 08:25:45 +00:00
- 异地备份目标：**Synology NAS** `103.40.14.100`，账号 `kts_bak`，落盘 **`/volume1/NetBackup/kaotings/`**（`db/` + `audio/`，**不在 home 下**），可用容量约 92 GiB。
- 传输方式：**tar over SSH**（端口 `52200`，**加密**），密钥认证（`/root/.ssh/id_ed25519_ktsbak`，公钥已加入 NAS `kts_bak` 的 `authorized_keys`）。**不使用** 公网明文 rsync 守护端口 `50873`。
- 为何不用 rsync：Synology 限制 rsync-over-SSH 仅能写入**已注册的共享文件夹**；`/volume1/NetBackup` 是普通目录（非 DSM 共享文件夹），rsync 会报 `0 bytes received`。而 `scp`/sftp 虽能写入，但会生成 **root 属主**文件，`kts_bak` 无法轮转删除。**tar over SSH（ssh exec）** 生成的文件属主为 `kts_bak`，可正常轮转。
- 升级路径：若在 DSM 中将 `NetBackup` 建为**共享文件夹**（控制面板→共享文件夹，授予 `kts_bak` 读写），即可改回 `rsync -a --delete` 获得增量传输。
docs: record configured offsite NAS backup (rsync over SSH) Offsite backup to the Synology NAS (103.40.14.100:52200) is now configured: pg_dump of business+CMS DBs (keep 7 per series) plus rsync of the audio dir, encrypted over SSH with key auth, on a daily 03:00 systemd timer. Document the destination, transport, script, timer, and restore-from-NAS procedure; mark the backup-location item resolved (retention/threshold/owner still pending).
2026-09-10 07:24:04 +00:00
- 脚本：`/usr/local/sbin/kaotings-backup.sh`（root 运行）：
1. `pg_dump --format=custom` 业务库（`/etc/kaotings/api.env` → `DATABASE_URL`）与 CMS 库（`/etc/kaotings/cms.env` → `DATABASE_URI`）。
docs: correct offsite backup to /volume1/NetBackup via tar-over-SSH Moved the backup out of the NAS home dir to /volume1/NetBackup/kaotings per request. Synology restricts rsync-over-SSH to registered shared folders and scp/sftp yields root-owned files that the backup account cannot rotate, so the backup now uses tar over SSH (kts_bak-owned, rotatable, encrypted, key auth). Document the transport rationale and the DSM shared-folder upgrade path for incremental rsync.
2026-09-10 08:25:45 +00:00
2. 每库本机保留最近 **7** 份（按序列独立轮转）。
3. `tar` over SSH 传 `db/` 与 `audio/`（`AUDIO_STORAGE_DIR`）到 NAS；再在远端按序列保留 7 份（ssh exec，可删除 `kts_bak` 属主文件）。
docs: record configured offsite NAS backup (rsync over SSH) Offsite backup to the Synology NAS (103.40.14.100:52200) is now configured: pg_dump of business+CMS DBs (keep 7 per series) plus rsync of the audio dir, encrypted over SSH with key auth, on a daily 03:00 systemd timer. Document the destination, transport, script, timer, and restore-from-NAS procedure; mark the backup-location item resolved (retention/threshold/owner still pending).
2026-09-10 07:24:04 +00:00
- 定时：systemd `kaotings-backup.timer`，每日 **03:00**（`Persistent=true`，`RandomizedDelaySec=30min`）触发 `kaotings-backup.service`。
docs: correct offsite backup to /volume1/NetBackup via tar-over-SSH Moved the backup out of the NAS home dir to /volume1/NetBackup/kaotings per request. Synology restricts rsync-over-SSH to registered shared folders and scp/sftp yields root-owned files that the backup account cannot rotate, so the backup now uses tar over SSH (kts_bak-owned, rotatable, encrypted, key auth). Document the transport rationale and the DSM shared-folder upgrade path for incremental rsync.
2026-09-10 08:25:45 +00:00
- 验证：备份已落盘 `/volume1/NetBackup/kaotings`（db 5+6 份、音频 22 文件，共 5.4M，属主 `kts_bak`）；`pg_restore --list` 确认归档可恢复。
docs: record configured offsite NAS backup (rsync over SSH) Offsite backup to the Synology NAS (103.40.14.100:52200) is now configured: pg_dump of business+CMS DBs (keep 7 per series) plus rsync of the audio dir, encrypted over SSH with key auth, on a daily 03:00 systemd timer. Document the destination, transport, script, timer, and restore-from-NAS procedure; mark the backup-location item resolved (retention/threshold/owner still pending).
2026-09-10 07:24:04 +00:00
docs: correct offsite backup to /volume1/NetBackup via tar-over-SSH Moved the backup out of the NAS home dir to /volume1/NetBackup/kaotings per request. Synology restricts rsync-over-SSH to registered shared folders and scp/sftp yields root-owned files that the backup account cannot rotate, so the backup now uses tar over SSH (kts_bak-owned, rotatable, encrypted, key auth). Document the transport rationale and the DSM shared-folder upgrade path for incremental rsync.
2026-09-10 08:25:45 +00:00
| 对象 | 工具 | 本机 | 异地（NAS `/volume1/NetBackup/kaotings`） |
docs: Phase 5 acceptance, deployment, operations and phase report Record the task/quota exception matrix (isolated fault injection) results, the admin quota concurrency fix, the isolated restore drill (business/CMS/audio, with the pre-existing missing-audio finding), upstream security launch blocker, and remaining minor items. Each item marked pass / fail / awaiting-external with test version and evidence. Do not mark reliability/launch-blocking items complete.
2026-09-09 21:01:27 +00:00
| --- | --- | --- | --- |
docs: correct offsite backup to /volume1/NetBackup via tar-over-SSH Moved the backup out of the NAS home dir to /volume1/NetBackup/kaotings per request. Synology restricts rsync-over-SSH to registered shared folders and scp/sftp yields root-owned files that the backup account cannot rotate, so the backup now uses tar over SSH (kts_bak-owned, rotatable, encrypted, key auth). Document the transport rationale and the DSM shared-folder upgrade path for incremental rsync.
2026-09-10 08:25:45 +00:00
| 业务库 `kaotings` | `pg_dump --format=custom` + tar | `/home/flym/backups/kaotings_<ts>.dump` | `…/kaotings/db/` |
| CMS 库 `kaotings_cms` | `pg_dump --format=custom` + tar | `/home/flym/backups/kaotings_cms_<ts>.dump` | `…/kaotings/db/` |
| 音频目录 | tar over SSH | `/home/flym/kaotings-audio/` | `…/kaotings/audio/` |
docs: Phase 5 acceptance, deployment, operations and phase report Record the task/quota exception matrix (isolated fault injection) results, the admin quota concurrency fix, the isolated restore drill (business/CMS/audio, with the pre-existing missing-audio finding), upstream security launch blocker, and remaining minor items. Each item marked pass / fail / awaiting-external with test version and evidence. Do not mark reliability/launch-blocking items complete.
2026-09-09 21:01:27 +00:00
docs: correct offsite backup to /volume1/NetBackup via tar-over-SSH Moved the backup out of the NAS home dir to /volume1/NetBackup/kaotings per request. Synology restricts rsync-over-SSH to registered shared folders and scp/sftp yields root-owned files that the backup account cannot rotate, so the backup now uses tar over SSH (kts_bak-owned, rotatable, encrypted, key auth). Document the transport rationale and the DSM shared-folder upgrade path for incremental rsync.
2026-09-10 08:25:45 +00:00
> 仍待用户确认：音频保留期、磁盘告警阈值、是否启用按 `expires_at` 的定时清理（见 §5）；备份保留份数当前=7（可调脚本 `KEEP`）。如希望增量传输，请在 DSM 将 `NetBackup` 建为共享文件夹（见上“升级路径”）。
docs: Phase 5 acceptance, deployment, operations and phase report Record the task/quota exception matrix (isolated fault injection) results, the admin quota concurrency fix, the isolated restore drill (business/CMS/audio, with the pre-existing missing-audio finding), upstream security launch blocker, and remaining minor items. Each item marked pass / fail / awaiting-external with test version and evidence. Do not mark reliability/launch-blocking items complete.
2026-09-09 21:01:27 +00:00
## 3. 容量 / 保留 / 清理（需用户决定的参数集中列出）
| 参数 | 当前值 | 需用户决定 |
| --- | --- | --- |
| 音频保留期 `AUDIO_RETENTION_SECONDS` | 7 天（默认） | 是（产品/隐私/存储预算） |
| 磁盘容量上限 / 告警阈值 | 根盘约 48 GiB 可用，无上限告警 | 是（阈值 + 通知方式） |
| 保留期清理策略 | 未启用定时清理（仅启动时清孤儿） | 是（是否按 `expires_at` 定时删除） |
| 原始文本保留期 | 存于 `tts_tasks.text` | 是（产品/隐私） |
docs: record configured offsite NAS backup (rsync over SSH) Offsite backup to the Synology NAS (103.40.14.100:52200) is now configured: pg_dump of business+CMS DBs (keep 7 per series) plus rsync of the audio dir, encrypted over SSH with key auth, on a daily 03:00 systemd timer. Document the destination, transport, script, timer, and restore-from-NAS procedure; mark the backup-location item resolved (retention/threshold/owner still pending).
2026-09-10 07:24:04 +00:00
| 备份保留份数 / 周期 | 已配置：每库 7 份 · 每日 03:00 · 异地 NAS | 否（可调脚本 `KEEP` 与 timer） |
docs: Phase 5 acceptance, deployment, operations and phase report Record the task/quota exception matrix (isolated fault injection) results, the admin quota concurrency fix, the isolated restore drill (business/CMS/audio, with the pre-existing missing-audio finding), upstream security launch blocker, and remaining minor items. Each item marked pass / fail / awaiting-external with test version and evidence. Do not mark reliability/launch-blocking items complete.
2026-09-09 21:01:27 +00:00
- 清理安全边界：
- 启动时 `cleanup_orphan_audio` 仅删除“**无** `available` 记录且 mtime>1h”的孤儿文件。
- 保留期清理在参数确认并实现后启用；**在此之前不删除任何既有音频**。
- 告警：磁盘使用率超阈值、`audio_files` 与磁盘文件数量长期不一致、Worker 长期无 `queued` 消化。
## 4. 恢复（含隔离演练结果）
### 4.1 流程
1. 停止写入（可选，视 RPO）：`systemctl stop kaotings-api`。
2. 建空库 → `pg_restore --clean --if-exists` 业务库 / CMS 库。
docs: correct offsite backup to /volume1/NetBackup via tar-over-SSH Moved the backup out of the NAS home dir to /volume1/NetBackup/kaotings per request. Synology restricts rsync-over-SSH to registered shared folders and scp/sftp yields root-owned files that the backup account cannot rotate, so the backup now uses tar over SSH (kts_bak-owned, rotatable, encrypted, key auth). Document the transport rationale and the DSM shared-folder upgrade path for incremental rsync.
2026-09-10 08:25:45 +00:00
3. 音频：从 NAS 用 `tar` over SSH 拉回 `/home/flym/kaotings-audio`（源：`/volume1/NetBackup/kaotings/audio/`）。
docs: Phase 5 acceptance, deployment, operations and phase report Record the task/quota exception matrix (isolated fault injection) results, the admin quota concurrency fix, the isolated restore drill (business/CMS/audio, with the pre-existing missing-audio finding), upstream security launch blocker, and remaining minor items. Each item marked pass / fail / awaiting-external with test version and evidence. Do not mark reliability/launch-blocking items complete.
2026-09-09 21:01:27 +00:00
4. 校验：行数、音频文件↔`audio_files` 关联、`checksum`、额度↔流水一致（见 §4.2 校验 SQL）。
5. `systemctl start kaotings-api`，`/healthz` 复核。
### 4.2 校验（隔离演练已执行）
- 业务库行数源=恢复：users 35 / quota 34 / tasks 28 / audio 26 / usage 61 / grants 10。
- 额度↔流水：`used` 与 `consume`、`reserved` 与 `reserve-consume-release` 均 0 不符。
- 音频：磁盘 22 文件全部被 `available` 引用（无孤儿）。
- **发现**：4 条 `status=available` 且任务 `succeeded` 的 `audio_files` 行在磁盘无文件（既有不一致，回放 404 降级）。待调查根因（见 §6）。
### 4.3 Worker 重启 / 额度对账
- 单 Worker（systemd 单进程），原子领取（`FOR UPDATE SKIP LOCKED`）+ 5 分钟租约 + 过期 `running` 回收（`WORKER_LEASE_EXPIRED`，释放冻结）。
- 对账：`GET /admin/usage/summary` 汇总 `quota_accounts` 与 `usage_records` 口径；`GET /admin/tts/tasks` 查结算状态（consumed/released/pending）。
- 不变式（每账户恒成立）：`limit_snapshot + adjustment - used - reserved >= 0`（由 Phase 5 异常矩阵 S3/S4 保证）。
## 5. 待用户确认清单（集中）
1. 音频保留期（`AUDIO_RETENTION_SECONDS`）。
2. 磁盘容量上限与告警阈值、通知方式。
3. 是否启用按 `expires_at` 的定时保留期清理。
4. 原始文本保留期。
docs: record configured offsite NAS backup (rsync over SSH) Offsite backup to the Synology NAS (103.40.14.100:52200) is now configured: pg_dump of business+CMS DBs (keep 7 per series) plus rsync of the audio dir, encrypted over SSH with key auth, on a daily 03:00 systemd timer. Document the destination, transport, script, timer, and restore-from-NAS procedure; mark the backup-location item resolved (retention/threshold/owner still pending).
2026-09-10 07:24:04 +00:00
5. 备份：异地 NAS 已配置（`103.40.14.100`，每日 03:00，每库 7 份）；**剩余待确认**：保留份数/周期是否调整、备份责任人。
docs: Phase 5 acceptance, deployment, operations and phase report Record the task/quota exception matrix (isolated fault injection) results, the admin quota concurrency fix, the isolated restore drill (business/CMS/audio, with the pre-existing missing-audio finding), upstream security launch blocker, and remaining minor items. Each item marked pass / fail / awaiting-external with test version and evidence. Do not mark reliability/launch-blocking items complete.
2026-09-09 21:01:27 +00:00
6. 上游 HTTPS/加密端点、允许来源网段、凭据轮换（见 `DEPLOYMENT.md` §4）。
## 6. 待调查 / 未闭环项（不标为完成）
- **4 条 `available` 音频行无对应文件**：既有 DB/文件系统不一致，待定位根因（文件被外部删除 or 写入路径异常）；恢复流程本身通过。
- **上游公网 HTTP**：上线阻断，待加密 + 来源限制 + 外部复测（`DEPLOYMENT.md` §4）。
- **音频容量/保留/清理参数**：待用户确认（§5）。
Reference in New Issue Copy Permalink