The repo today has a rich simulation cross-shard E2E (cross_shard_tx_sim) but
no dedicated production integration test that exercises a full cross-shard transaction on live wiring.
This module is a design-and-observability lab: you map the simulation test to production ingress, threading, and
multi-validator topology, then write acceptance criteria for a future test_production_cross_shard_*—the same spirit as
Single-Shard Production E2E hands-on, extended across shards.
In crates/simulation/tests/scenarios.rs, list (in a short markdown note or comment block you keep locally):
split_config() / ScenarioConfig sets (num_shards, validator count, connectivity).shard_for_node).completed, executed, finalized).This list becomes the checklist for production parity.
Search crates/production/tests/ for cross-shard or multi-shard coverage (today: typically none). Re-read
runner_boots_listens_and_shuts_down_cleanly from
single-shard production hands-on.
Write a table with three columns: Signal (e.g. listen address up, tx in mempool, cross-shard receipt), Single-shard prod test today, Needed for cross-shard prod E2E. Aim for at least six rows.
Without implementing a full new test yet, draft pseudocode or bullet steps for a future integration test:
ShardId) with production runner APIs.tx_submission_sender.shard-loop threads running, and a wall-clock bounded wait for finalization.Reference crates/paths from production harness analysis.
If you want a minimal repo change, add comment-only markers in
crates/production/tests/ (near the existing network runner test) pointing to this lab’s acceptance criteria—
or add a #[ignore] stub test named test_production_cross_shard_transaction with todo!() and your checklist in the doc comment.
Do not leave flaky or failing tests enabled in CI.
cargo test path.Continue to Performance measurement.