Skip to content

fix(pool): keep h2 protocol probe available after invalid request - #6207

Closed
mcollina wants to merge 1 commit into
mainfrom
fix/pool-invalid-first-request
Closed

mcollina wants to merge 1 commit into
mainfrom
fix/pool-invalid-first-request

Conversation

@mcollina

Copy link
Copy Markdown
Member

A request that fails validation before a client connects returns false from dispatch, which the pool currently treats as backpressure. That can leave an idle protocol-probe client marked busy and prevent later valid requests from reaching the origin.

Only mark a client busy when it is actually backpressured, including while draining queued requests. Add regression tests for an invalid first header or path followed by a valid request through Agent and Pool.

@codecov-commenter

codecov-commenter commented Nov 13, 2026 •

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 91.72%. Comparing base (6b22aba) to head (9c4a67f).

Additional details and impacted files
@@           Coverage Diff           @@
##             main    #6207   +/-   ##
=======================================
  Coverage   91.72%   91.72%           
=======================================
  Files         113      113           
  Lines       44835    44835           
=======================================
  Hits        41124    41124           
  Misses       3711     3711           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
  • 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

@mcollina
mcollina requested review from metcoder95 and ronag November 13, 2026 16:41
Signed-off-by: Matteo Collina <hello@matteocollina.com>
@mcollina
mcollina force-pushed the fix/pool-invalid-first-request branch from 690dc6f to 9c4a67f Compare November 20, 2026 18:10
@mcollina

Copy link
Copy Markdown
Member Author

Closing: #6180 fixed the root cause in DispatcherBase.dispatch(), which now returns true after a synchronous failure, so the pool no longer marks the client (or the HTTP/2 protocol probe) busy. The kBusy guards here became no-ops, and the tests added in #6180 already fail if that regresses.

@mcollina mcollina closed this Nov 20, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants