chore(apollo_l1_endpoint_monitor): add integration test

Author: Gilad Chase

Cargo.lock

@@ -13,7 +13,9 @@ tracing.workspace = true
 url = { workspace = true, features = ["serde"] }
 
 [dev-dependencies]
+alloy = { workspace = true, features = ["node-bindings"] }
 mockito.workspace = true
+papyrus_base_layer = { workspace = true, features = ["testing"] }
 tokio.workspace = true
 
 [lints]

crates/apollo_l1_endpoint_monitor/Cargo.toml

@@ -0,0 +1,69 @@
+use apollo_l1_endpoint_monitor::monitor::{
+    L1EndpointMonitor,
+    L1EndpointMonitorConfig,
+    L1EndpointMonitorError,
+};
+use papyrus_base_layer::test_utils::anvil;
+use url::Url;
+
+/// Integration test: two Anvil nodes plus a bogus endpoint to exercise cycling and failure.
+#[tokio::test]
+async fn end_to_end_cycle_and_recovery() {
+    // Spin up two ephemeral Anvil nodes.
+    // IMPORTANT: This is one of the only cases where two anvil nodes are needed simultaneously,
+    // since we are flow testing two separate L1 nodes. Other tests should never use more than one
+    // at a time!
+    let good_node_1 = anvil(None);
+    let good_url_1 = good_node_1.endpoint_url();
+    let good_node_2 = anvil(None);
+    let good_url_2 = good_node_2.endpoint_url();
+
+    // Bogus endpoint on port 1 that is likely to be unbound, see the unit tests for more details.
+    let bad_node_url = Url::parse("http://localhost:1").unwrap();
+
+    // Initialize monitor starting at the bad index.
+    let mut monitor = L1EndpointMonitor {
+        current_l1_endpoint_index: 0,
+        config: L1EndpointMonitorConfig {
+            ordered_l1_endpoint_urls: vec![
+                bad_node_url.clone(),
+                good_url_1.clone(),
+                good_url_2.clone(),
+            ],
+        },
+    };
+
+    // 1) First call: skip bad and take the first good one.
+    let active1 = monitor.get_active_l1_endpoint().await.unwrap();
+    assert_eq!(active1, good_url_1);
+    assert_eq!(monitor.current_l1_endpoint_index, 1);
+
+    // 2) Anvil 1 is going down.
+    drop(good_node_1);
+
+    // Next call: now the first good node is down, switch to second good node.
+    let active2 = monitor.get_active_l1_endpoint().await.unwrap();
+    assert_eq!(active2, good_url_2);
+    assert_eq!(monitor.current_l1_endpoint_index, 2);
+
+    // 3) Anvil 2 is now also down!
+    drop(good_node_2);
+
+    // All endpoints are now down --> error. Do this twice for idempotency.
+    for _ in 0..2 {
+        let result = monitor.get_active_l1_endpoint().await;
+        assert_eq!(result, Err(L1EndpointMonitorError::NoActiveL1Endpoint));
+        assert_eq!(monitor.current_l1_endpoint_index, 2);
+    }
+
+    // ANVIL node 1 has risen!
+    let good_node_1 = anvil(None);
+    // Anvil is configured to use an ephemeral port, so this new node will be bound to a fresh port.
+    // We cannot reuse the previous URL since the old port may no longer be available.
+    let good_url_1 = good_node_1.endpoint_url();
+    monitor.config.ordered_l1_endpoint_urls[1] = good_url_1.clone();
+    // Index wraps around from 2 to 0, 0 is still down so 1 is picked, which is operational now.
+    let active3 = monitor.get_active_l1_endpoint().await.unwrap();
+    assert_eq!(active3, good_url_1);
+    assert_eq!(monitor.current_l1_endpoint_index, 1);
+}

crates/apollo_l1_endpoint_monitor/tests/happy_flow.rs

@@ -100,9 +100,9 @@ pub fn get_test_ethereum_node() -> (TestEthereumNodeHandle, EthereumContractAddr
 
 // TODO(Arni): Make port non-optional.
 // Spin up Anvil instance, a local Ethereum node, dies when dropped.
-fn anvil(port: Option<u16>) -> AnvilInstance {
+pub fn anvil(port: Option<u16>) -> AnvilInstance {
     let mut anvil = Anvil::new();
-    // If the port is not set explicitly, a random value will be used.
+    // If the port is not set explicitly, a random ephemeral port is bound and used.
     if let Some(port) = port {
         anvil = anvil.port(port);
     }