feat: introduce load tests into devnet #664

wlawt · 2026-02-04T19:57:39Z

Supersedes #653

Motivation

Previously, there wasn't a way to easily test changes made across our services. We would deploy to the devnet and manually QA. As part of the production readiness of getting resource metering out, it's a requirement to be able to run the system locally and test locally first before rolling out changes.

Changes

This PR extends off of the system_tests which was originally inspired by scripts/devnet/ to include stress/load tests. It includes bringing in the Simulator.sol contract from base/benchmark and sending transactions that would be gas or time expensive (see storage slot modifications, calldata sizes)

Notes

When feat(builder): resource metering for transaction time limits #534 is merged, we can add/modify the load.rs to purposefully include transactions that would be rejected by the builder to test if the e2e behaviour works.

Co-authored-by: Niran Babalola niran.babalola@coinbase.com

cb-heimdall · 2026-02-04T19:57:42Z

🟡 Heimdall Review Status

Requirement Status More Info

Reviews

🟡 0/1

Denominator calculation

Show calculation

1 if user is bot 0

1 if user is external 0

2 if repo is sensitive 0

From .codeflow.yml 1

Additional review requirements

Show calculation

Max	0

0

From CODEOWNERS 0

Global minimum 0

Max 1

1

1 if commit is unverified 0

Sum 1

devnet/src/utils.rs

devnet/tests/load.rs

refcell · 2026-02-04T21:15:05Z

devnet/tests/load.rs

+fn assert_load_test_passed(stats: &Stats) {
+    assert!(
+        stats.submitted() >= 10,
+        "Expected at least 10 transactions to be submitted, got {}",
+        stats.submitted()
+    );
+    // TODO: once the resource metering changes are in, we should have different
+    // thresholds for failed transactions
+    assert!(stats.failed() <= 2, "Expected no more than 2 failed transactions, but {} failed", stats.failed());
+}


We should open up a ticket to track this. Generally would be nice to not commit TODOs into the codebase but at the minimum have tickets to track.

devnet/src/static/simulator_bytecode.hex

refcell

Few nits but otherwise good to merge

danyalprout · 2026-02-07T13:43:30Z

I think we should put load tests into base/infra. A few thoughts:

Our load testing framework should be able to run on a local devnet like this or via a CLI against a production network
It feels more like infra than protocol (i.e. load tests should run against proxyd/nodes as well as the protocol)

I think we need a bit of a plan to unify contender, this PR and gobrr. Lets chat on Monday and come up with a plan

spike

11b923b

wlawt requested a review from niran February 4, 2026 19:58