Add ballista-mvp part5 blog

systemxlabs · Apr 26, 2024 · 590bce1 · 590bce1
1 parent a985cdc
commit 590bce1
Show file tree

Hide file tree

Showing 3 changed files with 165 additions and 0 deletions.
diff --git a/content/blog/2024-04-26-ballista-mvp-part5/ballista-mvp-executor.drawio b/content/blog/2024-04-26-ballista-mvp-part5/ballista-mvp-executor.drawio
@@ -0,0 +1,112 @@
+<mxfile host="Electron" modified="2024-04-26T07:45:10.770Z" agent="Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) draw.io/24.2.5 Chrome/120.0.6099.109 Electron/28.1.0 Safari/537.36" etag="BrpBkrujaqnxWiCiN1Mv" version="24.2.5" type="device">
+  <diagram name="第 1 页" id="CD6N5ZRrbqDiyocHICmu">
+    <mxGraphModel dx="2901" dy="1196" grid="1" gridSize="10" guides="1" tooltips="1" connect="1" arrows="1" fold="1" page="1" pageScale="1" pageWidth="827" pageHeight="1169" math="0" shadow="0">
+      <root>
+        <mxCell id="0" />
+        <mxCell id="1" parent="0" />
+        <mxCell id="1erlsmTN53r9LNktAl-M-3" value="" style="rounded=1;whiteSpace=wrap;html=1;dashed=1;" vertex="1" parent="1">
+          <mxGeometry x="130" y="270" width="580" height="680" as="geometry" />
+        </mxCell>
+        <mxCell id="w1BsYSWlBb3kda8zS0cG-1" value="ExecutorGrpc API" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" parent="1" vertex="1">
+          <mxGeometry x="60" y="320" width="110" height="70" as="geometry" />
+        </mxCell>
+        <mxCell id="w1BsYSWlBb3kda8zS0cG-3" value="Flight API" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" parent="1" vertex="1">
+          <mxGeometry x="60" y="840" width="110" height="70" as="geometry" />
+        </mxCell>
+        <mxCell id="w1BsYSWlBb3kda8zS0cG-6" value="Task Runner Pool" style="ellipse;whiteSpace=wrap;html=1;fillColor=#e1d5e7;strokeColor=#9673a6;" parent="1" vertex="1">
+          <mxGeometry x="477" y="320" width="120" height="80" as="geometry" />
+        </mxCell>
+        <mxCell id="w1BsYSWlBb3kda8zS0cG-7" value="New Task Queue" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#fff2cc;strokeColor=#d6b656;" parent="1" vertex="1">
+          <mxGeometry x="250" y="340" width="150" height="40" as="geometry" />
+        </mxCell>
+        <mxCell id="w1BsYSWlBb3kda8zS0cG-8" value="Task Status Queue" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#fff2cc;strokeColor=#d6b656;" parent="1" vertex="1">
+          <mxGeometry x="512" y="450" width="50" height="160" as="geometry" />
+        </mxCell>
+        <mxCell id="w1BsYSWlBb3kda8zS0cG-10" value="" style="endArrow=classic;html=1;rounded=0;" parent="1" edge="1">
+          <mxGeometry width="50" height="50" relative="1" as="geometry">
+            <mxPoint x="190" y="360" as="sourcePoint" />
+            <mxPoint x="240" y="360" as="targetPoint" />
+          </mxGeometry>
+        </mxCell>
+        <mxCell id="w1BsYSWlBb3kda8zS0cG-11" value="" style="endArrow=classic;html=1;rounded=0;" parent="1" edge="1">
+          <mxGeometry width="50" height="50" relative="1" as="geometry">
+            <mxPoint x="417" y="359.5" as="sourcePoint" />
+            <mxPoint x="467" y="359.5" as="targetPoint" />
+          </mxGeometry>
+        </mxCell>
+        <mxCell id="w1BsYSWlBb3kda8zS0cG-14" value="" style="endArrow=classic;html=1;rounded=0;" parent="1" edge="1">
+          <mxGeometry width="50" height="50" relative="1" as="geometry">
+            <mxPoint x="537" y="410" as="sourcePoint" />
+            <mxPoint x="537" y="440" as="targetPoint" />
+          </mxGeometry>
+        </mxCell>
+        <mxCell id="w1BsYSWlBb3kda8zS0cG-15" value="Task Status Reporter" style="ellipse;whiteSpace=wrap;html=1;fillColor=#e1d5e7;strokeColor=#9673a6;" parent="1" vertex="1">
+          <mxGeometry x="477" y="660" width="120" height="80" as="geometry" />
+        </mxCell>
+        <mxCell id="w1BsYSWlBb3kda8zS0cG-16" value="" style="endArrow=classic;html=1;rounded=0;" parent="1" edge="1">
+          <mxGeometry width="50" height="50" relative="1" as="geometry">
+            <mxPoint x="538" y="620" as="sourcePoint" />
+            <mxPoint x="538" y="650" as="targetPoint" />
+          </mxGeometry>
+        </mxCell>
+        <mxCell id="w1BsYSWlBb3kda8zS0cG-17" value="Local Disk" style="rounded=1;whiteSpace=wrap;html=1;" parent="1" vertex="1">
+          <mxGeometry x="250" y="845" width="430" height="60" as="geometry" />
+        </mxCell>
+        <mxCell id="w1BsYSWlBb3kda8zS0cG-18" value="" style="endArrow=classic;startArrow=classic;html=1;rounded=0;" parent="1" edge="1">
+          <mxGeometry width="50" height="50" relative="1" as="geometry">
+            <mxPoint x="180" y="876" as="sourcePoint" />
+            <mxPoint x="240" y="876" as="targetPoint" />
+          </mxGeometry>
+        </mxCell>
+        <mxCell id="w1BsYSWlBb3kda8zS0cG-19" value="" style="endArrow=classic;html=1;rounded=0;" parent="1" edge="1">
+          <mxGeometry width="50" height="50" relative="1" as="geometry">
+            <mxPoint x="460" y="700" as="sourcePoint" />
+            <mxPoint x="40" y="700" as="targetPoint" />
+          </mxGeometry>
+        </mxCell>
+        <mxCell id="w1BsYSWlBb3kda8zS0cG-21" value="Heartbeater" style="ellipse;whiteSpace=wrap;html=1;fillColor=#e1d5e7;strokeColor=#9673a6;" parent="1" vertex="1">
+          <mxGeometry x="250" y="730" width="120" height="80" as="geometry" />
+        </mxCell>
+        <mxCell id="w1BsYSWlBb3kda8zS0cG-23" value="" style="endArrow=classic;html=1;rounded=0;" parent="1" edge="1">
+          <mxGeometry width="50" height="50" relative="1" as="geometry">
+            <mxPoint x="230" y="769.5" as="sourcePoint" />
+            <mxPoint x="40" y="770" as="targetPoint" />
+          </mxGeometry>
+        </mxCell>
+        <mxCell id="w1BsYSWlBb3kda8zS0cG-24" value="Scheduler" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;" parent="1" vertex="1">
+          <mxGeometry x="-90" y="310" width="100" height="490" as="geometry" />
+        </mxCell>
+        <mxCell id="w1BsYSWlBb3kda8zS0cG-25" value="" style="endArrow=classic;html=1;rounded=0;" parent="1" edge="1">
+          <mxGeometry width="50" height="50" relative="1" as="geometry">
+            <mxPoint x="19" y="359" as="sourcePoint" />
+            <mxPoint x="49" y="359" as="targetPoint" />
+          </mxGeometry>
+        </mxCell>
+        <mxCell id="w1BsYSWlBb3kda8zS0cG-26" value="Client" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;" parent="1" vertex="1">
+          <mxGeometry x="-85" y="830" width="90" height="40" as="geometry" />
+        </mxCell>
+        <mxCell id="w1BsYSWlBb3kda8zS0cG-27" value="Other Executors" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;" parent="1" vertex="1">
+          <mxGeometry x="-85" y="880" width="90" height="40" as="geometry" />
+        </mxCell>
+        <mxCell id="w1BsYSWlBb3kda8zS0cG-29" value="" style="endArrow=classic;startArrow=classic;html=1;rounded=0;" parent="1" edge="1">
+          <mxGeometry width="50" height="50" relative="1" as="geometry">
+            <mxPoint x="13" y="876" as="sourcePoint" />
+            <mxPoint x="53" y="876" as="targetPoint" />
+          </mxGeometry>
+        </mxCell>
+        <mxCell id="w1BsYSWlBb3kda8zS0cG-30" value="Executor" style="text;html=1;align=center;verticalAlign=middle;resizable=0;points=[];autosize=1;strokeColor=none;fillColor=none;fontStyle=1" parent="1" vertex="1">
+          <mxGeometry x="379" y="240" width="70" height="30" as="geometry" />
+        </mxCell>
+        <mxCell id="1erlsmTN53r9LNktAl-M-2" value="" style="endArrow=classic;html=1;rounded=0;" edge="1" parent="1">
+          <mxGeometry width="50" height="50" relative="1" as="geometry">
+            <mxPoint x="610" y="360" as="sourcePoint" />
+            <mxPoint x="640" y="830" as="targetPoint" />
+            <Array as="points">
+              <mxPoint x="640" y="360" />
+            </Array>
+          </mxGeometry>
+        </mxCell>
+      </root>
+    </mxGraphModel>
+  </diagram>
+</mxfile>
diff --git a/content/blog/2024-04-26-ballista-mvp-part5/ballista-mvp-executor.drawio.png b/content/blog/2024-04-26-ballista-mvp-part5/ballista-mvp-executor.drawio.png
diff --git a/content/blog/2024-04-26-ballista-mvp-part5/index.md b/content/blog/2024-04-26-ballista-mvp-part5/index.md
@@ -0,0 +1,53 @@
++++
+title = "Ballista 分布式查询引擎 - 计划执行"
+date = 2024-04-26
++++
+
+所有计划的执行均由 Executor 节点负责。
+
+![ballista-mvp-executor](./ballista-mvp-executor.drawio.png)
+
+Executor 节点启动时
+1. 启动 ExecutorGrpc 服务，负责接收 Scheduler 节点发送的任务和其他指令。
+2. 启动 Flight 服务，Client 和其他 Executor 通过 Flight 协议读取本机执行结果数据
+3. 向 Scheduler 注册
+4. 启动心跳任务，定时向所有 Scheduler 节点发送心跳
+5. 启动任务执行池，负责轮询队列获取新任务并执行
+6. 启动执行状态上报任务，负责上报任务状态给 Scheduler
+
+Executor 执行的基本单元是 Task，一个 Task 会执行一个 Job 内的一个 stage 对应的一个 partition。
+```rust
+pub struct TaskDefinition {
+    /// 在执行图中的唯一的（单调递增）
+    pub task_id: usize,
+    /// task 执行次数
+    pub task_attempt_num: usize,
+    /// 所属 job
+    pub job_id: String,
+    /// 所属 stage
+    pub stage_id: usize,
+    /// stage 执行次数
+    pub stage_attempt_num: usize,
+    /// partition
+    pub partition_id: usize,
+    /// stage 执行计划
+    pub plan: Arc<dyn ExecutionPlan>,
+    /// Scheduler 发起任务时间
+    pub launch_time: u64,
+    /// 所属会话
+    pub session_id: String,
+    /// 会话设置
+    pub props: Arc<HashMap<String, String>>,
+}
+```
+
+**Ballista 运行 CPU 密集型任务**
+
+Ballista 采用 [Tokio](https://github.com/tokio-rs/tokio) 作为 Rust 异步运行时。Tokio 使用协作式调度，异步任务会在执行到 .await 时切换其他任务执行，使用 Tokio 须遵守 `"Async code should never spend a long time without reaching an .await."`（参考 Tokio maintainer 的[这篇博客](https://ryhl.io/blog/async-what-is-blocking/)）。因此 CPU 密集型任务会长时间占用 CPU，导致运行时上的其他任务无法被及时执行（如心跳）。
+
+Ballista 解决此问题的办法是
+1. 采用单独的运行时专门跑 CPU 密集型任务
+2. IO 密集型任务运行时与 CPU 密集型任务运行时通过队列通信（创新新任务/任务执行结果通知）
+3. 当然 CPU 密集型任务代码仍要遵守 `"Async code should never spend a long time without reaching an .await."`
+
+更多参考 Datafusion maintainer 的这篇文章：[https://thenewstack.io/using-rustlangs-async-tokio-runtime-for-cpu-bound-tasks/](https://thenewstack.io/using-rustlangs-async-tokio-runtime-for-cpu-bound-tasks/) 。