13 days ago
I am evaluating a finite, nonroot cron worker with durable aggregate receipts. Before provisioning a synthetic pilot, can you clarify the supported interfaces below? No worker has been provisioned for this design, so these are platform-capability questions, not a live deployment failure.
-
What identifier can a container read that identifies one cron invocation and stays stable across automatic restarts? Deployment IDs are reused across cron runs. The public schema exposes deploymentInstanceExecutions IDs, but I have not established how a worker binds itself to one execution or how an observer distinguishes natural schedule from manual invocation. Which documented fields provide this binding and trigger provenance?
-
How can a UID/GID 10001 worker obtain a private writable volume without running the collector as root? Is explicit initial ownership/mode or a separately controlled initialization phase supported, preserving it through container replacement?
-
What supported read-only path retrieves a small receipt after a finite worker exits or crashes, without restarting it? CLI 5.29.0 volume-file selection requires an active deployment and uses SFTP into the instance. I have not tested live file access. Is offline access supported, or would a running helper be required?
-
Can an independent observer mount the worker volume read-only while the worker keeps its single-writer mount? If not, what transport and least-privilege authentication support receipt reads without broad project management, shell access or database credentials?
-
What minimum service requirements and compute/storage charges apply to the supported topology? I need a cost estimate before provisioning and independent missing-worker/missing-observer alerts.
Please do not create services, change permissions, issue credentials, start jobs or change networking. The proposed pilot would contain only synthetic data, with no production database or provider access. No paid commitment is requested.
1 Reply
13 days ago
For run tracking, railway cron list --json (mirroring the Cron Runs tab) returns each run with its own id, deploymentId, status (EXITED, CRASHED, RUNNING, SKIPPED and so on), createdAt, completedAt and meta, so an observer can tell runs apart even though they share a deployment ID. Off-schedule runs from railway cron run appear in the same list as scheduled ones. We can't confirm a run identifier readable inside the container or a field labelling manual versus scheduled runs, so the pilot shouldn't depend on either. See railway cron.
Volumes are mounted as root, and the documented approach for a non-root image is RAILWAY_RUN_UID=0. Volumes attach at container start, not during build or pre-deploy, so ownership setup belongs in the start command. See Using Volumes.
The supported file access path is the CLI (railway volume browse and railway volume files). We can't confirm offline reads of a stopped worker's volume or a second read-only mount, so plan a receipt path that needs neither.
Hobby is $5/month including $5 of usage. Compute is billed per minute only while the container runs, at $10/GB RAM/month and $20/vCPU/month, plus $0.15/GB/month volume storage and $0.05/GB egress. Builds are free, and Hobby volumes are capped at 5 GB. A short cron with a small volume usually stays within the included $5. See Pricing Plans. For alerting, railway cron list --json is scriptable to detect CRASHED or missing runs.
Status changed to Awaiting User Response Railway • 13 days ago
6 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 • 6 days ago