Skip to content

fix(docs): warn when worker-name is derived rather than pinned - #565

Merged
rubenvdlinde merged 1 commit into
mainfrom
fix/worker-name-derivation-warning
Aug 24, 2026
Merged

fix(docs): warn when worker-name is derived rather than pinned#565
rubenvdlinde merged 1 commit into
mainfrom
fix/worker-name-derivation-warning

Conversation

@rubenvdlinde

Copy link
Copy Markdown
Contributor

The trap I introduced in #555

#555 gave this workflow a worker-name default derived from cname — first label + -docs. That is only correct while the hostname and the worker still share a name.

Across the fleet they no longer do. Since the app-id rename:

host actually served by
integriq.conduction.nl openconnector-docs
dossiq.conduction.nl procest-docs
filinq.conduction.nl docudesk-docs
buildiq.conduction.nl openbuild-docs

…and so on for eight of twelve apps.

Why the default is dangerous rather than merely wrong

A derived name that does not exist is not an error. wrangler creates that worker and deploys to it, while the custom domains keep routing to the real one. The run goes green and the site never changes.

That is precisely the failure #555 was written to end — gh-pages written, nothing reading it — reintroduced one layer up by my own default.

Measured 2026-08-24: every fleet app had left worker-name unset, so every renamed one would have forked its deployment on its first successful publish. They now pin it explicitly (11 PRs).

What this changes

  • the input description leads with PIN IT and explains the failure mode
  • an unpinned run emits a warning naming the worker it derived and what happens if that worker is new

What it deliberately does not change

The default stays, rather than becoming required. A brand-new app whose host and worker do share a name is still the common case, and making it required would break every existing caller to guard against a condition only renamed apps have.

#555 gave this workflow a `worker-name` default derived from `cname` — first
label plus "-docs". That is only correct while the hostname and the worker still
share a name, and across the fleet they no longer do: integriq.conduction.nl is
served by `openconnector-docs`, dossiq.conduction.nl by `procest-docs`, and so
on for eight of twelve apps.

A derived name that does not exist is NOT an error. wrangler CREATES that worker
and deploys to it, while the custom domains keep routing to the real one — the
run goes green and the site never changes. That is the same failure this
workflow was written to end, reintroduced one layer up by my own default.

Measured 2026-08-24: every fleet app had left `worker-name` unset, so every
renamed one would have forked its deployment on the first successful publish.
They now pin it explicitly. This makes the trap loud for whoever comes next: the
input description says PIN IT and explains why, and an unpinned run emits a
warning naming the worker it derived and what will happen if that worker is new.

The default is kept rather than made required, because a site whose host and
worker do share a name is still the common case for a NEW app.
@rubenvdlinde
rubenvdlinde merged commit d1d7345 into main Aug 24, 2026
36 checks passed
@rubenvdlinde
rubenvdlinde deleted the fix/worker-name-derivation-warning branch August 24, 2026 10:16
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