a month ago
I am using Railway's connected integration in ChatGPT to set up a separate public-market-data collector. The existing API must remain unchanged.
Observed on 20 October 2026 around 06:52 UTC:
- create_service with staged=false and the production environment explicitly supplied returned success for aqlyra-spot-recorder.
-
- list_services includes it at project level.
-
- Six non-secret build/safety variables were accepted with skipDeploys=true, including RAILWAY_DOCKERFILE_PATH=collector/Dockerfile.
-
- update_service returns: Service instance not found.
-
- connect_service_source to our repository, collector branch and pinned commit returns success and says a deployment is starting.
-
- update_volume for the intended buffer reports successful attachment.
-
- describe_service still returns: Service has no configuration in this environment.
-
- list_deployments filtered to this service and production returns no deployments.
-
- describe_environment shows only the pre-existing services and unrelated volume, with no staged changes and no collector.
-
- The dashboard also does not show a deployed collector. An earlier collector object has the same symptoms.
Can you help determine whether this is an integration issue or a missing environment service instance, and how to establish the approved collector safely? We do not want more duplicate services, deletion of any resources/data, a plan upgrade, or changes to the existing API. We can provide project/service IDs privately if needed. No credentials are included in this report.
3 Replies
a month ago
On our side, the collector does not exist in your production environment yet. Production currently contains only the services that were there before, with no collector instance and no deployments for it. That matches the "Service instance not found" and "no configuration in this environment" responses. The calls that reported success, including the source connection and volume attachment, did not result in anything running in production.
Your workspace is also at its limit of 3 volumes per project. That limit will refuse a new volume for the collector until a volume slot is free.
So this is a missing environment service instance, not something wrong with your existing API. The supported way to create a service and apply it before it runs is in the dashboard: create an Empty Service with no source and click Deploy to apply it. Then set its variables, including RAILWAY_DOCKERFILE_PATH, and click Deploy again. Connect the repository and branch last, and that final Deploy starts its first build. Before each Deploy, check the environment's pending changes, because Deploy applies everything staged in that environment, not just the selected service.
Status changed to Awaiting User Response Railway • about 1 month ago
Railway
On our side, the collector does not exist in your production environment yet. Production currently contains only the services that were there before, with no collector instance and no deployments for it. That matches the "Service instance not found" and "no configuration in this environment" responses. The calls that reported success, including the source connection and volume attachment, did not result in anything running in production. Your workspace is also at its limit of 3 volumes per project. That limit will refuse a new volume for the collector until a volume slot is free. So this is a missing environment service instance, not something wrong with your existing API. The supported way to create a service and apply it before it runs is in the dashboard: create an Empty Service with no source and click Deploy to apply it. Then set its variables, including RAILWAY_DOCKERFILE_PATH, and click Deploy again. Connect the repository and branch last, and that final Deploy starts its first build. Before each Deploy, check the environment's pending changes, because Deploy applies everything staged in that environment, not just the selected service.
a month ago
Thank you. We inspected production again in both the dashboard and the integration before applying anything. Both currently show no pending changes. The five-change entry is in Activity under an earlier failed operation; its details include adding aqlyra-spot-collector, adding aqlyra-spot-buffer (requested 500 MB, iad), and mounting /data. Production inventory still exposes only creator-frozen64-evidence (256 MB) on the unrelated runner. The intended aqlyra-spot-buffer can be opened by its known volume URL, but Settings offers Mount to service, no associated region, and displays 1 MB; we cannot confirm that it is provisioned.
Before any Deploy, please identify the three volume records counted against the project limit, including which are provisioned, unattached, or left from failed operations, and explain how to reuse the intended buffer without creating another volume. Please also confirm whether any five-change patch is actually pending on your side. We have not deployed, deleted, wiped, resized, detached anything, or upgraded the plan during this reconciliation. Please do not delete or modify existing data/resources.
Status changed to Awaiting Railway Response Railway • about 1 month ago
a month ago
Nothing is pending in production right now. The most recent change there is marked as failed: it tried to mount the intended buffer at /data on the collector service, which has never been set up in production, so the mount did not land. The five-change entry in Activity belongs to that failed operation and is not waiting to be applied.
In production, the only volume in use is creator-frozen64-evidence, attached to the runner. The project is nonetheless at its limit of three volumes, and the intended buffer counts toward that limit even though it is not attached to anything. We can't confirm the buffer is usable as-is.
To reuse it without creating another volume, the collector needs to exist in production first. Create it as an empty service with no source, check that pending changes contain nothing else, and click Deploy to apply only that. Then use the buffer's Mount to service option to attach it to that service at /data and apply that change. Set variables and connect the repository only after the mount has applied. While the project is at its volume limit, creating a new volume instead of mounting this one will be refused.
Status changed to Awaiting User Response Railway • about 1 month ago
24 days ago
This thread has been marked as solved automatically due to a lack of recent activity. Please re-open this thread or create a new one if you require further assistance. Thank you!
Status changed to Solved Railway • 24 days ago