apollo_integration_tests: fix flaky test due to metric registration race

[AvivYossef-starkware]

Summary

• Fix flaky integration test failure where sequencer_num_accepted_txs panics when STATE_SYNC_PROCESSED_TRANSACTIONS metric isn't registered yet
• Return 0 instead of panicking, since this function is used in polling loops that retry

Context

This fixes a race condition observed in PR #14250 where the integration test failed with:

called `Result::unwrap()` on an `Err` value: MetricNotFound { metric_name: "apollo_state_sync_processed_transactions" }

The metric is registered asynchronously when StateSyncRunner::start() is called. The test can query the metric before registration completes. Since sequencer_num_accepted_txs is only used in run_until polling loops, returning 0 allows the test to wait for proper metric registration.

Test plan

• CI passes
• The fix is safe because all callers use this in retry loops that will eventually see the correct value once the metric is registered

Slack thread: https://starkwareindustries.slack.com/archives/C07E54Y3W3H/p1779980981138769?thread_ts=1779979080.466729&cid=C07E54Y3W3H

https://claude.ai/code/session_014UmiFHenEiE4vNnh8qbGkZ

---

Generated by Claude Code

[reviewable-StarkWare]

This change is 

[cursor]

PR Summary

Low Risk
Test-only changes in monitoring_utils; no production auth, sync, or transaction logic.

Overview
Hardens integration test monitoring helpers against metrics that are not registered yet when state sync or batcher components start asynchronously.

sequencer_num_accepted_txs now treats MetricNotFound for STATE_SYNC_PROCESSED_TRANSACTIONS as 0 instead of panicking, which matches its use in run_until polling until the metric appears. assert_no_reverted_txs retries until REVERTED_TRANSACTIONS is available, then asserts the count is zero.

^{Reviewed by Cursor Bugbot for commit a950914. Bugbot is set up for automated code reviews on this repo. Configure here.}

[dorimedini-starkware]

@dorimedini-starkware reviewed 1 file and all commit messages, and made 1 comment.
Reviewable status: all files reviewed, 1 unresolved discussion (waiting on AvivYossef-starkware).

---

crates/apollo_integration_tests/src/monitoring_utils.rs line 221 at r1 (raw file):

        "Sequencer {sequencer_idx} has {reverted_count} reverted transactions"
    );
}

wouldn't it be better to have a retry loop until the metric is registered? i.e. catch the specific "not registered" error and loop?
asking because the assert_no_reverted_txs can hit the same issue, but unwrap_or(0) is a bad idea there (false positive)

Code quote:

pub async fn sequencer_num_accepted_txs(monitoring_client: &MonitoringClient) -> usize {
    // If the sequencer accepted txs, sync should process them and update the respective metric.
    // Return 0 if the metric isn't registered yet (race with StateSyncRunner startup).
    monitoring_client
        .get_metric::<usize>(STATE_SYNC_PROCESSED_TRANSACTIONS.get_name())
        .await
        .unwrap_or(0)
}

pub async fn assert_no_reverted_txs(monitoring_client: &MonitoringClient, sequencer_idx: usize) {
    let reverted_count =
        monitoring_client.get_metric::<usize>(REVERTED_TRANSACTIONS.get_name()).await.unwrap();
    assert_eq!(
        reverted_count, 0,
        "Sequencer {sequencer_idx} has {reverted_count} reverted transactions"
    );
}

[AvivYossef-starkware]

@AvivYossef-starkware made 1 comment.
Reviewable status: all files reviewed, 1 unresolved discussion (waiting on claude).

---

crates/apollo_integration_tests/src/monitoring_utils.rs line 221 at r1 (raw file):

Previously, dorimedini-starkware wrote…

wouldn't it be better to have a retry loop until the metric is registered? i.e. catch the specific "not registered" error and loop?
asking because the assert_no_reverted_txs can hit the same issue, but unwrap_or(0) is a bad idea there (false positive)

@claude plz fix it according to Dori`s comment