fix(core): optimize warm cache performance for task execution

[FrozenPandaz]

Current Behavior

When running cached tasks, the task orchestrator processes each task individually: separate cache lookups, individual daemon IPC calls for output hash checking/recording, and per-task scheduling with a full array re-sort on each insert. For a workspace with 1,110 projects, this means ~1,100 individual JS→Rust→SQLite cache lookups, ~2,200 sequential daemon round-trips, and O(n² log n) scheduling overhead.

Expected Behavior

Warm cache runs should resolve quickly by batching all hot-path operations: cache lookups, daemon calls, scheduling, and filesystem scans.

Key Changes

Rust Native (cache.rs)

• NxCache.get_batch(): Single SQL UPDATE ... WHERE hash IN (...) RETURNING query + Rayon-parallel terminal output file reads. Replaces N individual JS→Rust boundary crossings and SQLite queries with 1.
• get_files_for_outputs_batch(): Rayon-parallel filesystem scanning for output expansion (from first commit).

Task Orchestrator (task-orchestrator.ts)

• Wire resolveCachedTasksBulk into coordinator loop: Bulk-resolves all cache hits before falling through to individual workers. Uses batched daemon calls for output hash checking.
• Reorder coordinator: bulk resolve before processTask: Skips N processTask lifecycle calls for tasks resolved from cache — only processes remaining cache misses.
• Remove dead executeDiscreteTaskLoop: Unused after coordinator introduction.

Task Scheduler (tasks-schedule.ts)

• Batch scheduling with single sort: Collect all schedulable roots, push them all at once, sort once. Previously re-sorted the entire array per task insert.

Cache (cache.ts)

• DbCache.getBatch(): TypeScript wrapper for the native batch cache lookup.

Daemon Outputs Tracking (outputs-tracking.ts)

• Skip filesystem scan when no recorded hash: Short-circuit outputsHashesMatch and outputsHashesMatchBatch when the daemon has no recorded hashes, avoiding unnecessary Rayon filesystem scans after nx reset.

From First Commit

• Batch daemon methods: recordOutputsHashBatch and outputsHashesMatchBatch
• Cache readProjectsConfigurationFromProjectGraph: Avoids rebuilding project map per task
• Skip recordOutputsHash for local-cache-kept-existing: Already has correct hashes
• Batch hashing per topological level

Benchmark Results (1,110 projects, warm cache)

Metric	Before	After	Improvement
build-warm	6.27s	2.12s	-66%

Related Issue(s)

[netlify]

✅ Deploy Preview for nx-docs ready!

Name	Link
🔨 Latest commit	0437035
🔍 Latest deploy log	https://app.netlify.com/projects/nx-docs/deploys/69d87dbba874270008157c16
😎 Deploy Preview	https://deploy-preview-35172--nx-docs.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[netlify]

✅ Deploy Preview for nx-dev ready!

Name	Link
🔨 Latest commit	0437035
🔍 Latest deploy log	https://app.netlify.com/projects/nx-dev/deploys/69d87dbba0b7a200080e8266
😎 Deploy Preview	https://deploy-preview-35172--nx-dev.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[nx-cloud]

View your CI Pipeline Execution ↗ for commit 0437035

Command	Status	Duration	Result
nx affected --targets=lint,test,build,e2e,e2e-c...	✅ Succeeded	51m 13s	View ↗
nx run-many -t check-imports check-lock-files c...	✅ Succeeded	4s	View ↗
nx-cloud record -- pnpm nx conformance:check	✅ Succeeded	7s	View ↗
nx build workspace-plugin	✅ Succeeded	<1s	View ↗
nx-cloud record -- nx format:check	✅ Succeeded	2s	View ↗
nx-cloud record -- nx sync:check	✅ Succeeded	<1s	View ↗

---

☁️ Nx Cloud last updated this comment at 2026-04-10 05:29:37 UTC

[nx-cloud]

Important

At least one additional CI pipeline execution has run since the conclusion below was written and it may no longer be applicable.

Nx Cloud is proposing a fix for your failed CI:

We added a .catch() handler on the fire-and-forget applyFromCacheOrRunTask call to capture errors (e.g. remote cache 401/connection failures from cache.put) that were previously silently swallowed by .finally(). The captured error is re-thrown after the coordinator loop exits, restoring the original behavior where these errors propagated to the CLI, printed the diagnostic message, and exited with a non-zero code.

Note

⏳ We are verifying this fix by re-running e2e-nx:e2e-ci--src/cache.test.ts.

Suggested Fix changes

diff --git a/packages/nx/src/tasks-runner/task-orchestrator.ts b/packages/nx/src/tasks-runner/task-orchestrator.ts
index 0c75521158..bbf924e089 100644
--- a/packages/nx/src/tasks-runner/task-orchestrator.ts
+++ b/packages/nx/src/tasks-runner/task-orchestrator.ts
@@ -227,6 +227,7 @@ export class TaskOrchestrator {
     parallelism: number
   ) {
     let inFlightWorkers = 0;
+    let firstWorkerError: unknown = null;
 
     while (true) {
       if (this.bailed || this.stopRequested) break;
@@ -287,8 +288,18 @@ export class TaskOrchestrator {
         dispatched = true;
         inFlightWorkers++;
         const groupId = this.closeGroup();
-        this.applyFromCacheOrRunTask(doNotSkipCache, task, groupId).finally(
-          () => {
+        this.applyFromCacheOrRunTask(doNotSkipCache, task, groupId)
+          .catch((e) => {
+            // Capture the first worker error so it can be re-thrown after
+            // the coordinator loop exits. This preserves the old behavior
+            // where errors (e.g. remote cache 401/connection failures) from
+            // cache.put propagated to crash the build with a visible message.
+            if (!firstWorkerError) firstWorkerError = e;
+            this.bailed = true;
+            this.waitingForTasks.forEach((f) => f(null));
+            this.waitingForTasks.length = 0;
+          })
+          .finally(() => {
             this.openGroup(groupId);
             inFlightWorkers--;
             // Wake coordinator — the decrement above may satisfy the
@@ -296,8 +307,7 @@ export class TaskOrchestrator {
             // when scheduleNextTasksAndReleaseThreads fired earlier.
             this.waitingForTasks.forEach((f) => f(null));
             this.waitingForTasks.length = 0;
-          }
-        );
+          });
       }
       if (dispatched) continue;
 
@@ -309,6 +319,11 @@ export class TaskOrchestrator {
       // 7. Wait for a worker to finish (woken by scheduleNextTasksAndReleaseThreads)
       await new Promise((res) => this.waitingForTasks.push(res));
     }
+
+    // Re-throw any error captured from a fire-and-forget worker so it
+    // propagates to run() and ultimately the CLI (which prints the message
+    // and exits with non-zero code).
+    if (firstWorkerError) throw firstWorkerError;
   }
 
   private async executeContinuousTaskLoop(continuousTaskCount: number) {

🔔 Heads up, your workspace has pending recommendations ↗ to auto-apply fixes for similar failures.

  

Or Apply changes locally with:

npx nx-cloud apply-locally cNYl-nDXD

Apply fix locally with your editor ↗   View interactive diff ↗

_{🎓 Learn more about Self-Healing CI on nx.dev}