The provision_tenant step is published and never consumed, leaving a tenant half bootstrapped
Table of Contents
This page is a capture in the inbox bucket of the product backlog — a pre-sprint idea, not yet pulled into a sprint as a story.
1. What
(One paragraph: the idea.)
2. Why
(Motivation, problem being solved, related context.)
3. References
4. See also
5. What the logs said
The step was published and nobody was listening, and the engine said so at the time:
[ERROR] [ores.workflow.service.workflow_engine] Step command could not be published to a stream, so no service was asked to act: workflow=f4e325d7-3c10-4ae0-ba43-cc3a55274d0c step=gleif.lei_entities.large subject=refdata.v1.ops.publish_legal_entities_from_dq error=JetStream publish failed: No responders available for request
The subject is new: it arrived with the legal entity work (75e227d3cf, merged as PR #4074), and the running refdata service was a binary built before it, so it had no handler and did not answer. Nothing was wrong with the workflow's routing.
Two defects follow, and neither is the stale binary:
- The engine knows the step reached no one. It logs an error and yet leaves the step `in_progress`, so the parent waits its full 20 minute timeout and the tenant stays in `bootstrapping`. A step that could not be published should fail at once, with the reason, and the parent should fail with it.
- The tenant is left in a state nothing can finish. `bootstrap setup` refuses because the system is no longer in bootstrap mode, and the workflow's own retry publishes to the same subject and waits again.