http: enable TCP keepalive on fetch sockets by cirospaciari · Pull Request #30627 · oven-sh/bun

cirospaciari · 2026-05-13T15:50:03Z

What does this PR do?

Enables TCP keepalive (SO_KEEPALIVE + TCP_KEEPIDLE=60s) on fetch() client sockets.

Without this, when a connection becomes half-open — the peer is gone but the FIN/RST never reached us (NAT timeout, wifi/cellular handoff, middlebox state eviction, VPN disconnect) — the kernel never discovers it. A streaming reader.read() on such a socket blocks forever (or until an application-level timeout).

Node's fetch (undici) sets SO_KEEPALIVE with TCP_KEEPIDLE=60s, so a half-open connection is detected at ~70s (60s idle + 10 probes × 1s). This makes Bun match that behavior via the existing socket.setKeepAlive() → bsd_socket_keepalive() path, which already hardcodes TCP_KEEPINTVL=1 and TCP_KEEPCNT=10.

The call is placed in onOpen() next to client.setTimeout(socket) (#30376) — socket-level, fires once per connection, inherited by keep-alive-reused requests.

How did you verify your code works?

Added test/js/web/fetch/fetch-tcp-keepalive.test.ts (Linux-only) that:

starts a streaming server, opens a fetch() to it
reads /proc/self/net/tcp and finds the client socket (ESTABLISHED, remote port = server port)
asserts the timer field is not 00:00000000 — i.e. the kernel's sk_timer (keepalive) is armed (timer_active=02)

Without this patch the timer field is 00; with it, 02:<jiffies>.

Bun's fetch() HTTP client does not set SO_KEEPALIVE on its sockets. When a connection becomes half-open — the peer closed but the FIN/RST never reached us (NAT timeout, wifi/cellular handoff, middlebox state eviction, VPN disconnect) — the kernel has no way to discover it without keepalive probes. A streaming reader.read() on such a socket blocks until an application-level timeout, if any. Node's fetch (undici) sets SO_KEEPALIVE with TCP_KEEPIDLE=60s, so a half-open connection is detected at ~70s (60s idle + 10 probes × 1s). This change makes Bun's fetch do the same, via the existing socket.setKeepAlive() → bsd_socket_keepalive() plumbing that already sets TCP_KEEPINTVL=1 and TCP_KEEPCNT=10. Placed in onOpen() alongside the existing client.setTimeout(socket) (#30376) — socket-level, fires once per connection, inherited by keep-alive-reused requests. The test reads /proc/self/net/tcp to verify the kernel's keepalive timer is armed on the client socket (timer_active=02). Linux-only.

robobun · 2026-05-13T15:50:15Z

^{Updated 9:57 AM PT - May 13th, 2026}

❌ @autofix-ci[bot], your commit d4fb0d0 has 2 failures in Build #54044 (All Failures):

🧪 To try this PR locally:

bunx bun-pr 30627

That installs a local version of the PR into your bun-30627 executable, so you can run:

bun-30627 --bun

coderabbitai · 2026-05-13T15:53:43Z

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro

Run ID: 0a5e2a86-499e-460f-b6a9-6a438a0fbced

📥 Commits

Reviewing files that changed from the base of the PR and between 6d5e33c and d4fb0d0.

📒 Files selected for processing (1)

test/js/web/fetch/fetch-tcp-keepalive.test.ts

Walkthrough

TCP keepalive is enabled on fetch client sockets with a 60-second idle timeout to detect stalled connections. A socket configuration change in onOpen() arms keepalive, and a Linux-only test validates the keepalive timer is active on established client sockets.

Changes

TCP Keepalive for Fetch

Layer / File(s)	Summary
TCP keepalive socket configuration `src/http/http.zig`	`onOpen()` calls `socket.setKeepAlive(true, 60)` after arming the idle timeout, enabling TCP keepalive detection on fetch client sockets.
TCP keepalive verification test `test/js/web/fetch/fetch-tcp-keepalive.test.ts`	Linux-only test spawns a streaming HTTP server, performs `fetch()`, inspects `/proc/self/net/tcp` for the client-side ESTABLISHED entry, and asserts the keepalive timer active field is `"02"` (armed).

🚥 Pre-merge checks | ✅ 4

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately and concisely summarizes the main change: enabling TCP keepalive on fetch sockets.
Description check	✅ Passed	The PR description covers both required sections: it clearly explains what the PR does (enables TCP keepalive with specific settings and reasoning) and how it was verified (with a detailed Linux test).
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

coderabbitai

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@test/js/web/fetch/fetch-tcp-keepalive.test.ts`:
- Around line 63-69: The assertion is too broad: instead of checking timerActive
is not "00", update the test to assert that timerActive specifically indicates
the keepalive timer code "02" (the PR notes "02:<jiffies>"). Locate the
timerActive check in fetch-tcp-keepalive.test.ts and replace the negative check
using timerActive with a positive check that timerActive begins with or matches
the "02" timer code (e.g., startsWith "02" or a regex matching the "02" prefix)
so the test only passes for the keepalive timer.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

Push a commit to this branch (recommended)
Create a new PR with the fixes

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro

Run ID: a0d3c677-476b-43fe-afce-edcd04cfab0e

📥 Commits

Reviewing files that changed from the base of the PR and between 3bf4b33 and fbb25f7.

📒 Files selected for processing (2)

src/http/http.zig
test/js/web/fetch/fetch-tcp-keepalive.test.ts

github-actions · 2026-05-13T15:55:31Z

Found 1 issue this PR may fix:

HTTPS requests hanging (regression 1.2.23 -> 1.3.x) #26066 - HTTPS requests hanging after sustained use (400-500 requests over minutes), consistent with half-open connections on pooled sockets where the peer disappeared without sending FIN/RST. TCP keepalive probes would detect these dead connections.

If this is helpful, copy the block below into the PR description to auto-close this issue on merge.

Fixes #26066

🤖 Generated with Claude Code

cirospaciari · 2026-05-13T15:57:14Z

@robobun adopt

robobun · 2026-05-13T15:59:56Z

✅ Adopted. Built locally, test passes with the fix and fails without it (timerActive is "00" on main, "02" with setKeepAlive(true, 60)).

Pushed 6d5e33c:

Tightened the assertion to toBe("02") per review (keepalive timer specifically, not just "any timer").
Replaced the sleep(50) race with await fetch() + reader.read() so the socket is guaranteed ESTABLISHED and the GET is ACKed before reading /proc/self/net/tcp.

CI status: the diff is green. The remaining red on build 54044 is pre-existing and unrelated:

test-http-should-emit-close-when-connection-is-aborted.ts — times out on Windows 2019 x64/x64-baseline. This same failure appears on 30+ unrelated PR builds (53900–54038: avif-linux, fake-timers, cluster-reuseport, Intl, etc.), none of which touch HTTP/fetch. main builds don't run the Windows 2019 lane at all, so it only surfaces on PRs. Not introduced here.
jsc-stress.test.ts (mprotect failed: 487) — JSC Wasm JIT failure on Windows, unrelated to HTTP.
dev-and-prod.test.ts — annotated flaky, HMR timing.

Skipping a retrigger since the Windows 2019 timeout is systematic across all PR builds and would recur. Ready for review/merge.

- Tighten assertion from `not.toBe("00")` to `toBe("02")` so only the keepalive timer (sk_timer) satisfies it, not a retransmit (01) or zero-window probe (04). - Await response headers + first chunk instead of `sleep(50)` so the socket is guaranteed ESTABLISHED and the outbound GET is ACKed before reading /proc/self/net/tcp — avoids a race where a retransmit timer would mask the keepalive timer. - Cancel the reader instead of draining the full body.

### What does this PR do? Ports the TCP keepalive call from `src/http/http.zig` `onOpen` (added in #30627, gated in #30640) into the Rust `on_open` in `src/http/lib.rs`. The Rust HTTP client landed in #30412 from a branch cut before #30627 merged, so the `set_keepalive` call was never carried over. This is the cause of `test/js/web/fetch/fetch-tcp-keepalive.test.ts` failing on every Linux test job since #30412 merged. The test reads `/proc/self/net/tcp` and expects `timer_active=02` (keepalive armed) but gets `00` because the Rust client never calls `set_keepalive`. Build #54331 (and every Linux job since): 2 pass / 2 fail; build #54099 (#30640's PR, Zig build): 4 pass / 0 fail; build #54202 (#30412's PR): test file not in tree, so it never ran. The change is the same shape as the Zig reference: after `self.set_timeout(socket)`, guarded by `!self.flags.disable_keepalive` so `fetch(url, { keepalive: false })` and `node:http` non-keepalive agents skip it (matching undici's `buildConnector`). `set_keep_alive` already exists on `NewSocketHandler` (`src/uws_sys/socket.rs:466`) and the `disable_keepalive` flag is already wired through. ### How did you verify your code works? I have not built or run this — opening it directly per request so CI validates. The covering test is `test/js/web/fetch/fetch-tcp-keepalive.test.ts` (Linux-only), which is the test currently failing on every PR.

coderabbitai Bot reviewed May 13, 2026

View reviewed changes

Comment thread test/js/web/fetch/fetch-tcp-keepalive.test.ts Outdated

claude Bot reviewed May 13, 2026

View reviewed changes

Comment thread test/js/web/fetch/fetch-tcp-keepalive.test.ts Outdated

robobun and others added 2 commits May 13, 2026 16:15

[autofix.ci] apply automated fixes

d4fb0d0

cirospaciari enabled auto-merge (squash) May 13, 2026 16:57

cirospaciari disabled auto-merge May 13, 2026 16:57

cirospaciari merged commit b9c757b into main May 13, 2026
75 of 77 checks passed

cirospaciari deleted the ciro/fetch-tcp-keepalive branch May 13, 2026 16:57

cirospaciari mentioned this pull request May 13, 2026

http: link undici reference for fetch TCP keepalive default #30637

Closed

robobun mentioned this pull request May 14, 2026

module-loader: narrow TLA re-entrancy skip to dynamic-import initiator (#30651) #30656

Closed

dylan-conway mentioned this pull request May 14, 2026

http: port fetch TCP keepalive to on_open in lib.rs #30733

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

http: enable TCP keepalive on fetch sockets#30627

http: enable TCP keepalive on fetch sockets#30627
cirospaciari merged 3 commits into
mainfrom
ciro/fetch-tcp-keepalive

cirospaciari commented May 13, 2026

Uh oh!

robobun commented May 13, 2026 •

edited

Loading

Uh oh!

coderabbitai Bot commented May 13, 2026 •

edited

Loading

Uh oh!

coderabbitai Bot left a comment

Uh oh!

Uh oh!

github-actions Bot commented May 13, 2026

Uh oh!

cirospaciari commented May 13, 2026

Uh oh!

Uh oh!

robobun commented May 13, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

cirospaciari commented May 13, 2026

What does this PR do?

How did you verify your code works?

Uh oh!

robobun commented May 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

coderabbitai Bot commented May 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Walkthrough

Changes

Uh oh!

coderabbitai Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

github-actions Bot commented May 13, 2026

Uh oh!

cirospaciari commented May 13, 2026

Uh oh!

Uh oh!

robobun commented May 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

robobun commented May 13, 2026 •

edited

Loading

coderabbitai Bot commented May 13, 2026 •

edited

Loading

robobun commented May 13, 2026 •

edited

Loading