fix(cloudflare): restore bounded probe scheduling - #3171
Conversation
commit: |
Performance benchmarksCompared 0 improved · 0 regressed · 6 within ±1.5%
View detailed results and traces 🟢 improvement · 🔴 regression · ⚫ change below 1.5% · paired base/head |
|
9ce781e to
dddf36f
Compare
dddf36f to
0b9049d
Compare
|
/bigbonk review for issues Please complete this review within 8 minutes. Review the exact current head and report only actionable in-scope findings. |
![ask-bonk[bot]](https://avatars.githubusercontent.com/in/2454580?s=60&v=4){kind=small}
| return; | ||
| return "done"; | ||
| } | ||
| if (result.state === "probe-failed" && result.retryable === true && attemptOptions.deferRetry) { |
There was a problem hiding this comment.
[P1] Do not reset the no-progress deadline for deferred failures
This branch is reached only after lastProgressAt was reset at line 628. Since every outer retry invokes classifyConcretePath again, repeated retryable responses keep extending warmCdnProbeTimeout even though no cacheability probe completed. For example, one always-503 target with phaseTimeoutMs: 25, retries: 60, and retryDelayMs: 10 runs through the retry budget instead of aborting after roughly 25ms. Move the progress update after this retry branch (or otherwise update it only for completed classifications) so the documented no-progress bound remains effective.
|
Posted 1 actionable inline finding. Focused tests passed. |
1 participant
Summary
Experimental top-of-stack change to test whether the cacheability probe coordinator caused the large-app performance regression.
Test plan