Repository navigation
Event-store replay: intended guarantees for engine-rejected fills? #5269
Replies: 1 comment 1 reply
|
Your numbers fit what the replay code does, though I only read the source and did not run your reproducer. In The engine already distinguishes the cases at runtime: the declined events are republished on a To narrow it down, compare the number of |
Uh oh!
There was an error while loading. Please reload this page.
Hi,
I’m evaluating the Rust event store for post-run execution reporting. I understand it is still early alpha, and I’d like to clarify the intended replay guarantees before building further on it.
An offline reproducer using
NautilusKernel, the real execution engine, and the default event-store capture produces this:allow_overfills=false.restore_cache_from_sealed_runsucceeds but restores a filled order and position of 7.This reproduces on develop commit
1974262d682bb690f7c476de646364f20fdfe148. The single-fill baseline and an explicitly allowed-overfill control both reconstruct correctly. There are no quantity updates or snapshot anchors involved.I saw the early-alpha caveat in #4128 and the discussion about deferring durable exactly-once application.
Is sealed-store cache reconstruction intended to reproduce the execution state the engine actually accepted? If so, is this discrepancy already known or part of planned event-store work, or would you prefer a separate bug report?
I have a small native integration-test reproducer and its output available.
All reactions