Merge pull request #94 from gadget-inc/parent-crash-monitoring

Make zombie children clean themselves up

Author: airhorns

.github/workflows/test.yml

@@ -13,7 +13,7 @@ jobs:
       - uses: actions/checkout@v2
       - uses: ./.github/actions/setup-test-env
       - run: pnpm build
-      - run: test/test.sh
+      - run: integration-test/test.sh
       - run: pnpm jest
 
   lint:

test/oom/package.json integration-test/oom/package.jsontest/oom/package.json renamed to integration-test/oom/package.json

@@ -0,0 +1,10 @@
+{
+  "name": "parent-crash",
+  "version": "1.0.0",
+  "description": "",
+  "main": "index.js",
+  "scripts": {
+    "test": "echo \"Error: no test specified\" && exit 1"
+  },
+  "license": "ISC"
+}

test/oom/run.ts integration-test/oom/run.tstest/oom/run.ts renamed to integration-test/oom/run.ts

@@ -0,0 +1,4 @@
+process.stderr.write("child started\n")
+setInterval(() => {
+  process.stderr.write("child still alive\n")
+}, 200)

test/oom/test.sh integration-test/oom/test.shtest/oom/test.sh renamed to integration-test/oom/test.sh

@@ -0,0 +1,50 @@
+#!/usr/bin/env zx
+const assert = require("assert");
+const { exec } = require("child_process");
+
+async function getChildPids(parentPid, callback) {
+  const result = await $`pgrep -P ${parentPid}`;
+  return result.stdout
+    .split("\n")
+    .filter((pid) => pid)
+    .map((pid) => parseInt(pid, 10));
+}
+
+function processIsRunning(pid) {
+  try {
+    process.kill(pid, 0);
+    return true;
+  } catch (e) {
+    return false;
+  }
+}
+
+const { setTimeout } = require("timers/promises");
+
+const main = async () => {
+  // launch the wds process
+  const parent = $`${__dirname}/../../pkg/wds.bin.js --watch ${__dirname}/run.ts`.nothrow();
+
+  // wait for the wds process to start
+  await setTimeout(500);
+
+  // get the pid of the child process that the parent wds supervisor will have started
+  const pids = await getChildPids(parent.child.pid);
+  assert(pids.length > 0, "no child pids found for supervisor process");
+
+  // SIGKILL the parent process, as if it OOMed or something like that to simulate a zombie child
+  console.log(`killing parent (${parent.child.pid})`);
+  await parent.kill(9);
+  assert.ok(processIsRunning(pids[0]), "test is broken, child process is not running immediately after parent is dead");
+
+  // ensure the children are dead too after their monitoring delay
+  await setTimeout(3000);
+
+  for (const pid of pids) {
+    assert.ok(!processIsRunning(pid), `child process ${pid} is still running after parent has been killed`);
+  }
+
+  await parent;
+};
+
+void main();

integration-test/parent-crash/package.json

@@ -0,0 +1,10 @@
+const http = require("http");
+
+const requestListener = function (req, res) {
+  res.writeHead(200);
+  res.end("Hey, Pluto!");
+};
+
+const server = http.createServer(requestListener);
+server.listen(8080);
+console.warn("Listening on 8080");