Event consumer queue-argument mismatch can terminate startup without containment #4

New issue

Open

opened 2026-09-06 19:12:31 +00:00 by nsaspy · 0 comments

nsaspy commented

2026-09-06 19:12:31 +00:00

Owner

Observed behavior

RabbitMQ logged PRECONDITION_FAILED for queue events: the declaration received no x-dead-letter-exchange while the existing queue had events.dead-letter. Event-consumer OPEN-STREAM then raised an unhandled CL-RABBIT:RABBITMQ-SERVER-ERROR, followed by server teardown errors.

Current source assessment (`21461ee`)

The loaded system order is consumers/consumers, owner-fixes, then retry-policy (source/starintel-gserver.asd:12-17). The active retry-policy open-stream redeclares queues without arguments (source/consumers/retry-policy.lisp:312-329), and its late create-rabbit-consumer replaces earlier definitions. start-event-consumer does not supply DLX topology (source/actor-systems/event-actor.lisp:240-267). The DLX-aware implementation in consumers/rabbit-settlement.lisp is not in the ASDF component list.

After the prior volume reset, the live events queue was observed with no DLX argument and two consumers. No current-source path declaring that same queue with DLX was found, so clean-start self-conflict is unproved. The confirmed defect is incompatible durable topology plus uncontained startup failure; the origin of the pre-existing DLX queue remains uncertain.

Expected behavior

Topology ownership/migration must be explicit and non-destructive. An incompatible non-empty durable queue must produce an actionable not-ready state rather than an unhandled worker exception/process teardown. Do not fix by disabling quarantine or deleting volumes.

Regression criteria

Test startup against existing queue definitions both with and without the supported arguments.
Prove the canonical declaration path and active load order.
Empty and non-empty incompatible queues have documented, safe outcomes.
Consumer startup failure is reflected in readiness and contained shutdown.

Research

STAR-RESEARCH-060 reassessment records the exact uncertainty and rejects the prior clean-start/stale-volume overclaim.

## Observed behavior RabbitMQ logged `PRECONDITION_FAILED` for queue `events`: the declaration received no `x-dead-letter-exchange` while the existing queue had `events.dead-letter`. Event-consumer `OPEN-STREAM` then raised an unhandled `CL-RABBIT:RABBITMQ-SERVER-ERROR`, followed by server teardown errors. ## Current source assessment (`21461ee`) The loaded system order is `consumers/consumers`, `owner-fixes`, then `retry-policy` (`source/starintel-gserver.asd:12-17`). The active `retry-policy` `open-stream` redeclares queues without arguments (`source/consumers/retry-policy.lisp:312-329`), and its late `create-rabbit-consumer` replaces earlier definitions. `start-event-consumer` does not supply DLX topology (`source/actor-systems/event-actor.lisp:240-267`). The DLX-aware implementation in `consumers/rabbit-settlement.lisp` is not in the ASDF component list. After the prior volume reset, the live `events` queue was observed with no DLX argument and two consumers. No current-source path declaring that same queue with DLX was found, so clean-start self-conflict is unproved. The confirmed defect is incompatible durable topology plus uncontained startup failure; the origin of the pre-existing DLX queue remains uncertain. ## Expected behavior Topology ownership/migration must be explicit and non-destructive. An incompatible non-empty durable queue must produce an actionable not-ready state rather than an unhandled worker exception/process teardown. Do not fix by disabling quarantine or deleting volumes. ## Regression criteria - Test startup against existing queue definitions both with and without the supported arguments. - Prove the canonical declaration path and active load order. - Empty and non-empty incompatible queues have documented, safe outcomes. - Consumer startup failure is reflected in readiness and contained shutdown. ## Research STAR-RESEARCH-060 reassessment records the exact uncertainty and rejects the prior clean-start/stale-volume overclaim.

No labels

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

starintel-labs/starintel-server#4

No description provided.

Rows
Columns