Skip to content

Check local write fairness at receiver flush boundaries - #12

Closed
jeregrine wants to merge 1 commit into
test-receiver-cluster-isolationfrom
test-receiver-fairness
Closed

Check local write fairness at receiver flush boundaries#12
jeregrine wants to merge 1 commit into
test-receiver-cluster-isolationfrom
test-receiver-fairness

Conversation

@jeregrine

Copy link
Copy Markdown
Member

Problem

Observing how much replication backlog remains when a local caller wakes up makes fairness assertions depend on scheduler timing. A local write can also complete eventually while still waiting behind the entire finite backlog, masking a missing fairness turn.

Fix

Generate receiver buffer sizes, incoming batch sizes, backlog lengths, and local register/join operations for both registry and PG receiver lanes.

Suspend one shard, enqueue the receiver backlog, and explicitly wait until the public local request is queued before resuming. Assert that the local event appears at the first receiver flush and before the backlog finishes. Check the remaining remote event order, the exact local event, settled contents, and index consistency.

Supporting information

Stacked on #11 in the follow-up chain rooted at #8. This PR adds only fairness properties and documentation, reusing the existing fixture and history oracle.

The property controls a finite backlog on one shard. It does not claim a deterministic BEAM scheduler or fairness under an infinite stream of control traffic. Production behavior is unchanged.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant