shishengliang/lingniu-vehicle-ingest
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

refactor(go): make raw topics the realtime source

Browse Source
This commit is contained in:
lingniu
2026-07-03 08:25:14 +08:00
parent c2d058bf75
commit 4c34c1221b
14 changed files with 151 additions and 44 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
docs/go-version-data-flow.html
View File

@@ -380,7 +380,7 @@
</div>
<div class="node bus">
<strong>Subjects</strong>
<p><code>vehicle.raw.go.gb32960.v1</code><br><code>vehicle.raw.go.jt808.v1</code><br><code>vehicle.raw.go.yutong-mqtt.v1</code><br><code>vehicle.event.go.unified.v1</code></p>
<p><code>vehicle.raw.go.gb32960.v1</code><br><code>vehicle.raw.go.jt808.v1</code><br><code>vehicle.raw.go.yutong-mqtt.v1</code></p>
</div>
<div class="node warning">
<strong>Async + Retry</strong>
@@ -396,7 +396,7 @@
</div>
<div class="node bus">
<strong>Kafka Topics</strong>
<p>与 NATS subject 同名：raw 三个协议 + unified 一个实时主题。</p>
<p>与 NATS subject 同名：三个协议 RAW topic。<code>vehicle.event.go.unified.v1</code> 仅作为显式兼容开关保留。</p>
</div>
<div class="node warning">
<strong>失败语义</strong>
@@ -416,7 +416,7 @@
</div>
<div class="node store">
<strong>realtime-api 消费器</strong>
<p>消费 unified topic，写 Redis 在线状态、实时合并快照、各协议 realtime-raw。</p>
<p>消费三个 RAW topic，写 Redis 在线状态、实时合并快照、各协议 realtime-raw。</p>
</div>
<div class="node api">
<strong>HTTP API</strong>
@@ -433,7 +433,7 @@
<div class="mini-step"><b>2. 拆包</b><span>TCP 按协议提取完整帧；MQTT 直接按消息处理。</span></div>
<div class="mini-step"><b>3. 解析</b><span>生成 parsed 全量结构化数据和 fields 核心字段。</span></div>
<div class="mini-step"><b>4. 绑定 VIN</b><span>根据 VIN/phone/device_id/plate 生成 vehicle_key。</span></div>
<div class="mini-step"><b>5. 发 NATS</b><span>写 RAW subject；同时写 unified subject。</span></div>
<div class="mini-step"><b>5. 发 NATS</b><span>写协议 RAW subject；unified 只在兼容开关打开时写。</span></div>
<div class="mini-step"><b>6. 桥接 Kafka</b><span>bridge 写 Kafka 成功后 ACK NATS。</span></div>
<div class="mini-step"><b>7. 多路消费</b><span>历史、实时、统计各自消费 Kafka，互不阻塞。</span></div>
</div>