Skip to content

Probe a real outbound fetch alongside the local probes - #1673

Merged
RhysSullivan merged 1 commit into
mainfrom
diag/fetch-probe
Aug 18, 2026
Merged

Probe a real outbound fetch alongside the local probes#1673
RhysSullivan merged 1 commit into
mainfrom
diag/fetch-probe

Conversation

@RhysSullivan

Copy link
Copy Markdown
Collaborator

The previous probes came back: preWorkMs 0-4ms, cacheProbeMs 2-21ms, timerProbeMs 1ms — on requests whose wall time was 4000-7800ms.

So there is no startup/queue penalty, and local I/O (Cache API, timers) is fast. Both of those probes stay inside the isolate. The one class not yet measured is a real outbound network subrequest, which is exactly what the docs proxy implicates: executor.mintlify.dev answers in 0.098s directly and 3-6s through the Worker.

This times a fetch to a small unrelated endpoint on every request. If fetchProbeMs is seconds, outbound subrequests are the cost and the search narrows to why.

Cache and timer probes come back in single-digit ms while the same
requests take 4-6s wall, so local I/O is not the cost. Neither probe
leaves the isolate. Time a real outbound subrequest, the class the docs
proxy implicates (0.098s direct vs 3-6s through the Worker).
@RhysSullivan
RhysSullivan merged commit fde4450 into main Aug 18, 2026
@RhysSullivan
RhysSullivan deleted the diag/fetch-probe branch August 18, 2026 22:04
@github-actions

github-actions Bot commented Aug 18, 2026

Copy link
Copy Markdown
Contributor

Cloudflare preview

Torn down — the PR is closed.

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.

1 participant