fix(wrangler): keep `wrangler dev` alive when a proxied request to the UserWorker fails transiently - #15252

GregoryCollett · 2026-08-18T15:28:38Z

Fixes #14926. Also mitigates the empty-message symptom of #14906 for this path by including the request method, URL, attempt count and underlying exception in the logged error.

The bug

wrangler dev runs the ProxyWorker (binding the public port) and the UserWorker runtime as separate workerd processes. When a proxied request to the UserWorker fails while the UserWorker's origin is unchanged, the ProxyWorker reports {type: "error"}, ProxyController.onProxyWorkerMessage escalates it via emitErrorEvent, DevEnv.handleErrorEvent falls through to the fatal re-emit, and the entire dev server exits with an empty ✘ [ERROR] (the error crosses a JSON boundary, so castErrorCause produces a message-less Error), leaving the port unbound. In CI test suites, one transient failure kills every remaining test with ERR_CONNECTION_REFUSED — several independent reports on #14926 describe exactly this.

Root cause of the transient failure itself

kj HTTP defaults collide: the client pool closes idle connections after 5s (HttpClientSettings.idleTimeout) and the server closes idle keep-alive connections after 5s (HttpServerSettings.pipelineTimeout). For a pooled ProxyWorker→UserWorker connection idling ≈5s, both timers fire ~simultaneously; a request written into the connection as the server's close is in flight gets an RST, surfaced as Error: Network connection lost..

Reproduced organically (no fault injection): bursts of ~48 concurrent GETs separated by idle gaps swept 3.0→7.9s in 0.1s steps against wrangler dev hit the fatal within one sweep, at a 4.2s outer gap (per-connection idle 4.2–5.2s given the ~1s burst spread) — failed requests interleaved with same-millisecond 200s (stale pool checkouts die, fresh connections succeed), while the UserWorker runtime never exited and its isolate was never re-created. Deterministic variant: SIGKILL the UserWorker runtime under request load (miniflare restarts it reusing the port, so the in-flight failures are same-origin).

The fix (two small pieces)

ProxyWorker.ts: same-origin fetch failures now retry bodyless (GET/HEAD) requests — immediately, then once more after 250ms — before reporting. The retry draws a fresh connection, absorbing the race. This mirrors the existing requeue-on-reload behaviour and its rationale comment ("it would be incorrect to retry non-idempotent requests"); non-GET/HEAD behaviour is unchanged. Recovered requests log a console.warn so absorbed events remain visible. The reported error now includes method, URL and attempt count.
DevEnv.ts: handleErrorEvent classifies "Error inside ProxyWorker" reports as recoverable — logged via logger.error with the underlying exception — instead of falling through to the fatal top-level re-emit. This sits alongside the existing recoverable carve-out for other ProxyController reasons. One failed proxied request should never take down the dev session.

With both pieces, the kill-the-runtime repro goes from "dev server exits, port unbound" to: in-flight GETs retry across the restart (5 of 11 recovered in my run), the rest fail individually with the error logged, and the server keeps serving.

Tests
- Tests included/updated — DevEnv.test.ts gains a test asserting a ProxyWorker error report is logged and does not re-emit a fatal top-level error event.
- Automated tests not possible - manual testing has been completed as follows: the retry path was validated against a real wrangler dev session two ways — (a) the organic idle-gap sweep described above (7,200 requests: zero server-side failures with the fix; the fatal within one sweep without it), and (b) SIGKILL-ing the UserWorker runtime under load (server survives; GETs recover across the restart). Note: I could not run the wrangler vitest suite locally because @cloudflare/remote-bindings#build fails in my environment on a pristine main checkout (its tsdown embed-workers plugin resolves template paths incorrectly there), so I'm relying on CI for the full suite; turbo check:type --filter=wrangler and oxfmt pass locally.
Public documentation
- Documentation not necessary because: behavioural bug fix in wrangler dev internals; a changeset is included.

Note

This is a contribution from an AI agent: Claude Code (Claude Fable 5), working under the direction and review of Gregory Collett.

…UserWorker fails transiently When a request proxied to the UserWorker failed while the UserWorker's origin was unchanged — most commonly a reused keep-alive connection the UserWorker's HTTP server closed at the same moment the request was written to it (kj's client-pool idleTimeout and server pipelineTimeout both default to 5s, so a connection idling ~5s races the close) — the ProxyWorker reported a fatal error and the whole dev server exited with an empty error message, leaving the port unbound. The ProxyWorker now retries bodyless (GET/HEAD) requests before reporting, absorbing the transient failure on a fresh connection, and DevEnv classifies an exhausted or non-retriable report as recoverable: it is logged with the request method, URL, attempt count and underlying exception, and the dev server keeps serving. Only the affected request fails. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

changeset-bot · 2026-08-18T15:28:47Z

🦋 Changeset detected

Latest commit: fc9c3f1

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 3 packages

Name	Type
wrangler	Patch
@cloudflare/vite-plugin	Patch
@cloudflare/vitest-pool-workers	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

workers-devprod · 2026-08-18T15:29:11Z

Codeowners approval required for this PR:

@cloudflare/wrangler

Show detailed file reviewers

.changeset/tidy-moons-provide.md: [@cloudflare/wrangler]
packages/wrangler/src/tests/api/startDevWorker/DevEnv.test.ts: [@cloudflare/wrangler]
packages/wrangler/src/api/startDevWorker/DevEnv.ts: [@cloudflare/wrangler]
packages/wrangler/templates/startDevWorker/ProxyWorker.ts: [@cloudflare/wrangler]

pkg-pr-new · 2026-08-18T15:32:19Z

@cloudflare/autoconfig

npm i https://pkg.pr.new/@cloudflare/autoconfig@15252

@cloudflare/build-output-utils

npm i https://pkg.pr.new/@cloudflare/build-output-utils@15252

@cloudflare/codemods

npm i https://pkg.pr.new/@cloudflare/codemods@15252

@cloudflare/config

npm i https://pkg.pr.new/@cloudflare/config@15252

create-cloudflare

npm i https://pkg.pr.new/create-cloudflare@15252

@cloudflare/deploy-helpers

npm i https://pkg.pr.new/@cloudflare/deploy-helpers@15252

@cloudflare/kv-asset-handler

npm i https://pkg.pr.new/@cloudflare/kv-asset-handler@15252

miniflare

npm i https://pkg.pr.new/miniflare@15252

@cloudflare/pages-functions

npm i https://pkg.pr.new/@cloudflare/pages-functions@15252

@cloudflare/pages-shared

npm i https://pkg.pr.new/@cloudflare/pages-shared@15252

@cloudflare/unenv-preset

npm i https://pkg.pr.new/@cloudflare/unenv-preset@15252

@cloudflare/vite-plugin

npm i https://pkg.pr.new/@cloudflare/vite-plugin@15252

@cloudflare/vitest-plugin

npm i https://pkg.pr.new/@cloudflare/vitest-plugin@15252

@cloudflare/workers-auth

npm i https://pkg.pr.new/@cloudflare/workers-auth@15252

@cloudflare/workers-editor-shared

npm i https://pkg.pr.new/@cloudflare/workers-editor-shared@15252

@cloudflare/workers-utils

npm i https://pkg.pr.new/@cloudflare/workers-utils@15252

wrangler

npm i https://pkg.pr.new/wrangler@15252

commit: fc9c3f1

dario-piotrowicz

Thank you very much @GregoryCollett 😄

The PR looks overall good to me but there are a few bits that I think should be looked at before merging, could you please have a look? 🙏

- Move the GET/HEAD retry into fetch's rejection handler so errors thrown while post-processing a received response are reported, never retried (and can never re-run the UserWorker's handler). - Make the attempt parameter optional and 0-indexed. - Report exhausted retries explicitly ("failed after 3 attempts"). - data: {} in the DevEnv test; full issue URL in the DevEnv comment. - Rewrite the changeset to be user-facing. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

GregoryCollett · 2026-08-20T15:54:54Z

All review comments addressed in 8adab46:

Retry is attached to fetch's rejection handler, so only connection-level failures with no received response are retried; post-processing errors report on the first attempt and can never re-run the UserWorker's handler.
Exhausted retries are explicit in the logged error ((failed after 3 attempts)).
attemptUserWorkerFetch(attempt = 0) with a bare call, plus a JSDoc (kept as an arrow — the handlers need the ProxyWorker's lexical this to re-read this.proxyData across async boundaries).
Changeset rewritten user-facing; data: {} in the test; full issue URL in the DevEnv comment.

Re-verified end-to-end on the updated branch: SIGKILLing the UserWorker runtime under a 4-thread GET storm — 1034 requests, 0 client-visible failures, 4 recovered on the delayed third attempt, dev server kept serving. DevEnv.test.ts also passes locally now — the local-vitest blocker in the PR description turned out to be Node 22.14: below 22.18, tsdown bundles its config file, which rewrites import.meta.dirname inside scripts/embed-workers.ts to the package root, so ../templates resolves to packages/templates and @cloudflare/remote-bindings#build fails. On Node ≥ 22.18 it builds fine. Happy to file that separately.

While in here I noticed packages/remote-bindings/templates/startDevWorker/ProxyWorker.ts — the trimmed fork used for remote proxy sessions — still has the pre-fix error path. Happy to open a follow-up issue/PR if useful.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

The ProxyWorker's report crosses a JSON boundary, so castErrorCause wraps a plain object in a message-less Error and the real detail (method, URL, attempt count, underlying error) sits on .cause. Reading event.cause.message logged nothing after the colon — the same empty-error symptom this PR fixes. The test now builds its cause through JSON and castErrorCause, so it can no longer pass with a message the production path never has. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

workers-devprod · 2026-09-02T11:59:07Z

Codeowners approval required for this PR:

✅ @cloudflare/wrangler

Show detailed file reviewers

A retry scheduled before a reload begins would fire with the captured proxyData, so it could reach the pre-reload Worker or send a stale preview token. Re-check proxyData identity when the timer fires and requeue instead, letting processQueue rebuild the request from current routing and headers. Compares identity rather than origin because the origin can be unchanged while the headers are stale. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

workers-devprod

Codeowners reviews satisfied

dario-piotrowicz · 2026-09-07T12:12:22Z

Thanks for the changes @GregoryCollett 🙏

The rejected UserWorker fetch is now answered with a 502 inside the ProxyWorker itself, so the "Error inside ProxyWorker" report can only mean the ProxyWorker failed while post-processing a received response — a genuine proxy defect. Drop the recoverable branch DevEnv gained in cloudflare#15252 so such reports are fatal again, and flip its regression test to assert the re-emit. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RBs8hK2eBFkHBgik22QB97

GregoryCollett requested a review from workers-devprod as a code owner August 18, 2026 15:28

workers-devprod added this to workers-sdk Aug 18, 2026

github-project-automation Bot moved this to Untriaged in workers-sdk Aug 18, 2026

workers-devprod requested review from a team and dario-piotrowicz and removed request for a team August 18, 2026 15:29

This comment was marked as resolved.

Sign in to view

dario-piotrowicz reviewed Aug 18, 2026

View reviewed changes

report the true attempt count on proxied-request errors

5fc41f0

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

This comment was marked as resolved.

Sign in to view

SrivatsaRv mentioned this pull request Aug 24, 2026

Enforce Air-Runtime Test Layers in Make and Change-Aware CI SrivatsaRv/vector-engagements-labs#63

Open

16 tasks

jeffwall-curlewlabs mentioned this pull request Sep 2, 2026

Regression of #4562: Error inside ProxyWorker kills wrangler dev, and the printed error is empty #15317

Open

This comment was marked as resolved.

Sign in to view

GregoryCollett requested a review from dario-piotrowicz September 2, 2026 12:45

julioarguello mentioned this pull request Sep 3, 2026

dx(stack): wrangler dev dies mid-run with an empty-message proxy error julioarguello/www-serverstartup-io#390

Closed

dario-piotrowicz approved these changes Sep 7, 2026

View reviewed changes

workers-devprod approved these changes Sep 7, 2026

View reviewed changes

github-project-automation Bot moved this from Untriaged to Approved in workers-sdk Sep 7, 2026

dario-piotrowicz merged commit 682cd44 into cloudflare:main Sep 7, 2026
88 of 92 checks passed

github-project-automation Bot moved this from Approved to Done in workers-sdk Sep 7, 2026

workers-devprod mentioned this pull request Sep 7, 2026

Version Packages #15510

Merged

kody-bot mentioned this pull request Sep 7, 2026

Cleanup: drop wrangler ProxyWorker patch after #15207/#15252 ship kentcdodds/kody#1835

Closed

This was referenced Sep 8, 2026

build(deps): bump eslint, turbo, playwright, rolldown and the routine catalog simshanith/lit-ui-router#760

Merged

build(deps): bump wrangler to 4.129.1 simshanith/lit-ui-router#763

Merged

tbuckley mentioned this pull request Sep 9, 2026

Keep Wrangler dev proxy alive during polling tbuckley/tines#156

Merged

exKAZUu mentioned this pull request Sep 15, 2026

[wrangler] Log ProxyWorker request error details at debug level #15207

Merged

5 tasks

nocoo mentioned this pull request Sep 23, 2026

chore(deps): apply safe dependency updates nocoo/zhe#944

Merged

This was referenced Sep 28, 2026

The hosted flows' remote cursor checks fail when the machine is heavily loaded ndelangen/dunezone#1343

Open

Wrangler 4.139.0 keeps the local dev server up when an idle proxy connection drops ndelangen/dunezone#1450

Merged

Conversation

GregoryCollett commented Aug 18, 2026 • edited by devin-ai-integration Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

The bug

Root cause of the transient failure itself

The fix (two small pieces)

Uh oh!

changeset-bot Bot commented Aug 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🦋 Changeset detected

Uh oh!

workers-devprod commented Aug 18, 2026

Uh oh!

pkg-pr-new Bot commented Aug 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This comment was marked as resolved.

Uh oh!

dario-piotrowicz left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

GregoryCollett commented Aug 20, 2026

Uh oh!

This comment was marked as resolved.

Uh oh!

workers-devprod commented Sep 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This comment was marked as resolved.

Uh oh!

workers-devprod left a comment

Choose a reason for hiding this comment

Uh oh!

dario-piotrowicz commented Sep 7, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

GregoryCollett commented Aug 18, 2026 •

edited by devin-ai-integration Bot

Loading

changeset-bot Bot commented Aug 18, 2026 •

edited

Loading

pkg-pr-new Bot commented Aug 18, 2026 •

edited

Loading

workers-devprod commented Sep 2, 2026 •

edited

Loading