fix(web-probe): make observe control commands short-lived (#707)

Co-authored-by: Codex <codex@noreply.local>

scripts/src/hwlab-node-impl.ts

@@ -72,6 +72,7 @@ interface NodeWebProbeObserveOptions {
   viewport: string;
   sampleIntervalMs: number;
   screenshotIntervalMs: number;
+  observerRefreshIntervalMs: number;
   maxSamples: number;
   commandTimeoutSeconds: number;
   waitMs: number;
@@ -6900,6 +6901,7 @@ function parseNodeWebProbeObserveOptions(
     "--viewport",
     "--sample-interval-ms",
     "--screenshot-interval-ms",
+    "--observer-refresh-interval-ms",
     "--max-samples",
     "--command-timeout-seconds",
     "--wait-ms",
@@ -6940,9 +6942,10 @@ function parseNodeWebProbeObserveOptions(
     viewport: optionValue(args, "--viewport") ?? "1440x900",
     sampleIntervalMs: positiveIntegerOption(args, "--sample-interval-ms", 5000, 600000),
     screenshotIntervalMs: positiveIntegerOption(args, "--screenshot-interval-ms", 300000, 86_400_000),
+    observerRefreshIntervalMs: positiveIntegerOption(args, "--observer-refresh-interval-ms", 180000, 86_400_000),
     maxSamples: positiveIntegerOption(args, "--max-samples", 0, 10_000_000),
     commandTimeoutSeconds: positiveIntegerOption(args, "--command-timeout-seconds", 55, 3600),
-    waitMs: positiveIntegerOption(args, "--wait-ms", observeActionRaw === "command" || observeActionRaw === "stop" ? 30000 : 0, 600000),
+    waitMs: positiveIntegerOption(args, "--wait-ms", 0, 600000),
     tailLines: positiveIntegerOption(args, "--tail-lines", 5, 200),
     maxFiles: positiveIntegerOption(args, "--max-files", 80, 5000),
     collectFile,
@@ -7405,6 +7408,20 @@ function runNodeWebProbeObserveStart(
   const stateDir = options.stateDir ?? defaultStateDir;
   const runnerB64 = Buffer.from(nodeWebObserveRunnerSource(), "utf8").toString("base64");
   const webProbeProxy = nodeWebProbeHostProxyEnv(spec);
+  const runnerEnvAssignments = [
+    ...webProbeProxy.envAssignments,
+    `HWLAB_WEB_BASE_URL=${shellQuote(options.url)}`,
+    `HWLAB_WEB_USER=${shellQuote(secretSpec.bootstrapAdminUsername)}`,
+    `HWLAB_WEB_PASS=${shellQuote(material.password)}`,
+    `UNIDESK_WEB_OBSERVE_STATE_DIR=${shellQuote(stateDir)}`,
+    `UNIDESK_WEB_OBSERVE_JOB_ID=${shellQuote(jobId)}`,
+    `UNIDESK_WEB_OBSERVE_TARGET_PATH=${shellQuote(options.targetPath)}`,
+    `UNIDESK_WEB_OBSERVE_SAMPLE_INTERVAL_MS=${shellQuote(String(options.sampleIntervalMs))}`,
+    `UNIDESK_WEB_OBSERVE_SCREENSHOT_INTERVAL_MS=${shellQuote(String(options.screenshotIntervalMs))}`,
+    `UNIDESK_WEB_OBSERVE_OBSERVER_REFRESH_INTERVAL_MS=${shellQuote(String(options.observerRefreshIntervalMs))}`,
+    `UNIDESK_WEB_OBSERVE_MAX_SAMPLES=${shellQuote(String(options.maxSamples))}`,
+    `UNIDESK_WEB_OBSERVE_VIEWPORT=${shellQuote(options.viewport)}`,
+  ].join(" ");
   const script = [
     "set -eu",
     `state_dir=${shellQuote(stateDir)}`,
@@ -7413,20 +7430,7 @@ function runNodeWebProbeObserveStart(
     "runner=\"$state_dir/observer-runner.mjs\"",
     `node -e "require('fs').writeFileSync(process.argv[1], Buffer.from(process.argv[2], 'base64'))" "$runner" ${shellQuote(runnerB64)}`,
     "chmod 700 \"$runner\"",
-    [
+    `if command -v setsid >/dev/null 2>&1; then setsid env ${runnerEnvAssignments} node "$runner" >"$state_dir/stdout.log" 2>"$state_dir/stderr.log" </dev/null & else nohup env ${runnerEnvAssignments} node "$runner" >"$state_dir/stdout.log" 2>"$state_dir/stderr.log" </dev/null & fi`,
-      ...webProbeProxy.envAssignments,
-      `HWLAB_WEB_BASE_URL=${shellQuote(options.url)}`,
-      `HWLAB_WEB_USER=${shellQuote(secretSpec.bootstrapAdminUsername)}`,
-      `HWLAB_WEB_PASS=${shellQuote(material.password)}`,
-      `UNIDESK_WEB_OBSERVE_STATE_DIR=${shellQuote(stateDir)}`,
-      `UNIDESK_WEB_OBSERVE_JOB_ID=${shellQuote(jobId)}`,
-      `UNIDESK_WEB_OBSERVE_TARGET_PATH=${shellQuote(options.targetPath)}`,
-      `UNIDESK_WEB_OBSERVE_SAMPLE_INTERVAL_MS=${shellQuote(String(options.sampleIntervalMs))}`,
-      `UNIDESK_WEB_OBSERVE_SCREENSHOT_INTERVAL_MS=${shellQuote(String(options.screenshotIntervalMs))}`,
-      `UNIDESK_WEB_OBSERVE_MAX_SAMPLES=${shellQuote(String(options.maxSamples))}`,
-      `UNIDESK_WEB_OBSERVE_VIEWPORT=${shellQuote(options.viewport)}`,
-      "nohup node \"$runner\" >\"$state_dir/stdout.log\" 2>\"$state_dir/stderr.log\" </dev/null &",
-    ].join(" "),
     "pid=$!",
     "printf '%s\\n' \"$pid\" >\"$state_dir/pid\"",
     "sleep 1",
@@ -7527,6 +7531,12 @@ function renderWebObserveStatusTable(value: Record<string, unknown>): string {
   const result = record(value.result);
   const samples = Array.isArray(tails?.samples) ? tails.samples.map(record).filter((item): item is Record<string, unknown> => item !== null).slice(-5) : [];
   const controls = Array.isArray(tails?.control) ? tails.control.map(record).filter((item): item is Record<string, unknown> => item !== null).slice(-5) : [];
+  const completedControlIds = new Set(controls
+    .filter((item) => item.phase === "completed" || item.phase === "failed")
+    .map((item) => webObserveText(item.commandId))
+    .filter((item) => item !== "-"));
+  const activeControl = controls.slice().reverse().find((item) => item.phase === "started" && !completedControlIds.has(webObserveText(item.commandId))) ?? null;
+  const activeControlAgeSeconds = activeControl === null ? null : Math.max(0, Math.round((Date.now() - Date.parse(webObserveText(activeControl.ts))) / 1000));
   const network = Array.isArray(tails?.network)
     ? tails.network.map(record).filter((item): item is Record<string, unknown> => item !== null).filter((item) => {
       const status = typeof item.status === "number" ? item.status : null;
@@ -7575,6 +7585,17 @@ function renderWebObserveStatusTable(value: Record<string, unknown>): string {
       ]]),
       "",
     ] : []),
+    ...(activeControl !== null ? [
+      "Active command:",
+      webObserveTable(["TYPE", "COMMAND", "AGE_S", "STARTED_AT", "DETAIL"], [[
+        webObserveText(activeControl.type),
+        webObserveShort(webObserveText(activeControl.commandId), 28),
+        webObserveText(activeControlAgeSeconds),
+        webObserveShort(webObserveText(activeControl.ts), 24),
+        webObserveShort(webObserveText(activeControl.detail), 80),
+      ]]),
+      "",
+    ] : []),
     "Recent samples:",
     webObserveTable(["SEQ", "TS", "PATH", "ROUTE_SESSION", "ACTIVE_SESSION", "MSG", "TRACE"], samples.length > 0 ? samples.map((sample) => [
       webObserveText(sample.seq),