Since the custom *.ai.studio URL became mandatory, deploying a newly created app
fails at the URL step. The UI shows only “deployment failed” with no reason.
Remixing an existing app and deploying that works reliably — same account, same
region, same day.
Evidence from Cloud Run audit logs (project gen-lang-client-*):
Failing case
- The URL availability check (CreateDomainMapping with dryRun=all) runs
repeatedly, but CreateService is never called. No deploy request is ever issued. - One user made 31 dry-run calls across 4 sessions over 3 days, with zero
CreateService. - Names returning NOT_FOUND (“Resource ‘’ of kind ‘ROUTE’ … does not exist”),
which appears to indicate the name IS available, still did not let the deploy
proceed.
Working case
- Remix deploys succeed normally: CreateService, then CreateDomainMapping with
spec.routeName populated, then Ready. - In every success we observed, the gap between the last dry-run and the actual
create was 13-20 seconds.
One notable case: a user deployed a new app successfully (Service Ready,
DomainMapping Ready + DomainRoutable), and 37 seconds later both were deleted
from the UI. Four further attempts all stalled at the dry-run stage.
Region is not the factor. We initially suspected it, but failures occur in
us-west1 (a domain-mapping-supported region) and successes occur in asia-east1
on the same day with the same account.
Requests
- Surface the actual failure reason in the UI instead of a bare
“deployment failed”. - Fix the new-app path so it reaches CreateService, or document the sequence
that works. - Consider allowing users to skip the custom URL, as was possible before.
Side note: the custom URL validator rejects the tokens “qa”, “test” and “region”
when hyphen-separated (“region-valid” rejected, “regionvalid” accepted). We could
not find this documented anywhere, and it cost users many retries.