Files
foxhunt/tli/tests/e2e/train_stop_e2e_test.rs
jgrusewski 7458f1be01 feat(wave12): E2E validation complete - 225-feature pipeline ready
 Validation Results:
- PPO training: 24.2s (1 epoch, 950 samples, dim=225)
- Feature extraction: 105μs/bar (9.5x faster than target)
- Model checkpoint: 293KB (147KB actor + 146KB critic)
- GPU memory: 145MB used (96.4% headroom)
- Zero dimension mismatches

📊 Success Criteria (5/5):
 Feature dimension = 225 (Wave C 201 + Wave D 24)
 Model state_dim = 225
 Training completed without errors
 Checkpoint saved successfully
 No dimension mismatch errors

📁 Training Data Ready:
- ES.FUT: 2.9MB, 180 days
- NQ.FUT: 4.4MB, 180 days
- 6E.FUT: 2.8MB, 180 days
- ZN.FUT: 65KB, 90 days (clean)

🚀 Next: Full production model retraining (4 models, ~10min GPU time)

🤖 Generated with Claude Code (https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-10-22 22:48:04 +02:00

405 lines
11 KiB
Rust

//! End-to-End Tests for `tli train stop` Command
//!
//! Tests stopping training jobs and cleanup
use tli::proto::ml_training::{
ml_training_service_server::MlTrainingServiceServer, TrainingStatus,
ml_training_service_client::MlTrainingServiceClient, StopTrainingRequest,
};
use tonic::transport::{Server, Channel};
use std::net::SocketAddr;
use tokio::time::Duration;
use super::mock_ml_training_service::{MockMlTrainingService, MockTrainingState};
use super::test_fixtures;
/// Helper to start mock gRPC server
async fn start_mock_server(state: MockTrainingState) -> SocketAddr {
let addr: SocketAddr = "127.0.0.1:0".parse().unwrap();
let service = MockMlTrainingService::new(state);
let listener = tokio::net::TcpListener::bind(addr).await.unwrap();
let local_addr = listener.local_addr().unwrap();
tokio::spawn(async move {
Server::builder()
.add_service(MlTrainingServiceServer::new(service))
.serve_with_incoming(tokio_stream::wrappers::TcpListenerStream::new(listener))
.await
.unwrap();
});
tokio::time::sleep(Duration::from_millis(100)).await;
local_addr
}
/// TEST 1: Stop running job
#[tokio::test]
async fn test_stop_running_job() {
let state = MockTrainingState::default();
let job = test_fixtures::create_test_job(
"stop_test_1",
"TFT",
TrainingStatus::Running,
);
state.add_job(job);
let addr = start_mock_server(state.clone()).await;
let channel = Channel::from_shared(format!("http://{}", addr))
.unwrap()
.connect()
.await
.unwrap();
let mut client = MlTrainingServiceClient::new(channel);
let request = StopTrainingRequest {
job_id: "stop_test_1".to_string(),
reason: "User requested stop".to_string(),
};
let response = client.stop_training(request).await;
assert!(response.is_ok(), "Should successfully stop running job");
let result = response.unwrap().into_inner();
assert!(result.success);
// Verify job status changed to Stopped
let job = state.get_job("stop_test_1").unwrap();
assert_eq!(job.status, TrainingStatus::Stopped as i32);
}
/// TEST 2: Stop entire batch
#[tokio::test]
async fn test_stop_batch() {
let state = MockTrainingState::default();
// Create batch jobs
for i in 0..4 {
let job = test_fixtures::create_test_job(
&format!("batch_job_{}", i),
"TFT",
TrainingStatus::Running,
);
state.add_job(job);
}
let addr = start_mock_server(state.clone()).await;
let channel = Channel::from_shared(format!("http://{}", addr))
.unwrap()
.connect()
.await
.unwrap();
let mut client = MlTrainingServiceClient::new(channel);
// Stop each job in batch
for i in 0..4 {
let request = StopTrainingRequest {
job_id: format!("batch_job_{}", i),
reason: "Batch stop".to_string(),
};
let response = client.stop_training(request).await;
assert!(response.is_ok(), "Should stop batch job {}", i);
}
// Verify all jobs stopped
for i in 0..4 {
let job = state.get_job(&format!("batch_job_{}", i)).unwrap();
assert_eq!(job.status, TrainingStatus::Stopped as i32);
}
}
/// TEST 3: Stop already completed job → Error
#[tokio::test]
async fn test_stop_completed_job() {
let state = MockTrainingState::default();
let job = test_fixtures::create_test_job(
"completed_job",
"DQN",
TrainingStatus::Completed,
);
state.add_job(job);
let addr = start_mock_server(state).await;
let channel = Channel::from_shared(format!("http://{}", addr))
.unwrap()
.connect()
.await
.unwrap();
let mut client = MlTrainingServiceClient::new(channel);
let request = StopTrainingRequest {
job_id: "completed_job".to_string(),
reason: "Test stop completed".to_string(),
};
// Mock allows this, but real implementation might error
let response = client.stop_training(request).await;
assert!(response.is_ok());
}
/// TEST 4: Force stop (cleanup)
#[tokio::test]
async fn test_force_stop() {
let state = MockTrainingState::default();
let job = test_fixtures::create_test_job(
"force_stop_test",
"PPO",
TrainingStatus::Running,
);
state.add_job(job);
let addr = start_mock_server(state.clone()).await;
let channel = Channel::from_shared(format!("http://{}", addr))
.unwrap()
.connect()
.await
.unwrap();
let mut client = MlTrainingServiceClient::new(channel);
let request = StopTrainingRequest {
job_id: "force_stop_test".to_string(),
reason: "Force stop with cleanup".to_string(),
};
let response = client.stop_training(request).await;
assert!(response.is_ok());
let job = state.get_job("force_stop_test").unwrap();
assert_eq!(job.status, TrainingStatus::Stopped as i32);
}
/// TEST 5: Stop non-existent job → 404
#[tokio::test]
async fn test_stop_nonexistent_job() {
let state = MockTrainingState::default();
let addr = start_mock_server(state).await;
let channel = Channel::from_shared(format!("http://{}", addr))
.unwrap()
.connect()
.await
.unwrap();
let mut client = MlTrainingServiceClient::new(channel);
let request = StopTrainingRequest {
job_id: "nonexistent_job_123".to_string(),
reason: "Test".to_string(),
};
let response = client.stop_training(request).await;
assert!(response.is_err(), "Should return 404 for non-existent job");
assert!(response.unwrap_err().message().contains("not found"));
}
/// TEST 6: Idempotent stop
#[tokio::test]
async fn test_idempotent_stop() {
let state = MockTrainingState::default();
let job = test_fixtures::create_test_job(
"idempotent_test",
"MAMBA_2",
TrainingStatus::Running,
);
state.add_job(job);
let addr = start_mock_server(state.clone()).await;
let channel = Channel::from_shared(format!("http://{}", addr))
.unwrap()
.connect()
.await
.unwrap();
let mut client = MlTrainingServiceClient::new(channel);
let request1 = StopTrainingRequest {
job_id: "idempotent_test".to_string(),
reason: "First stop".to_string(),
};
let response1 = client.stop_training(request1).await;
assert!(response1.is_ok());
// Stop again - should be idempotent
let request2 = StopTrainingRequest {
job_id: "idempotent_test".to_string(),
reason: "Second stop".to_string(),
};
let response2 = client.stop_training(request2).await;
assert!(response2.is_ok(), "Stop should be idempotent");
}
/// TEST 7: Stop with custom reason
#[tokio::test]
async fn test_stop_with_reason() {
let state = MockTrainingState::default();
let job = test_fixtures::create_test_job(
"reason_test",
"TFT",
TrainingStatus::Running,
);
state.add_job(job);
let addr = start_mock_server(state).await;
let channel = Channel::from_shared(format!("http://{}", addr))
.unwrap()
.connect()
.await
.unwrap();
let mut client = MlTrainingServiceClient::new(channel);
let request = StopTrainingRequest {
job_id: "reason_test".to_string(),
reason: "Low validation accuracy".to_string(),
};
let response = client.stop_training(request).await;
assert!(response.is_ok());
let result = response.unwrap().into_inner();
assert!(result.success);
assert!(result.message.contains("stopped"));
}
/// TEST 8: Concurrent stop requests
#[tokio::test]
async fn test_concurrent_stops() {
let state = MockTrainingState::default();
// Create multiple jobs
for i in 0..10 {
let job = test_fixtures::create_test_job(
&format!("concurrent_stop_{}", i),
"DQN",
TrainingStatus::Running,
);
state.add_job(job);
}
let addr = start_mock_server(state.clone()).await;
// Stop all concurrently
let mut handles = vec![];
for i in 0..10 {
let addr_clone = addr;
let handle = tokio::spawn(async move {
let channel = Channel::from_shared(format!("http://{}", addr_clone))
.unwrap()
.connect()
.await
.unwrap();
let mut client = MlTrainingServiceClient::new(channel);
let request = StopTrainingRequest {
job_id: format!("concurrent_stop_{}", i),
reason: "Concurrent stop test".to_string(),
};
client.stop_training(request).await
});
handles.push(handle);
}
// Wait for all stops
for handle in handles {
let result = handle.await.unwrap();
assert!(result.is_ok());
}
// Verify all stopped
for i in 0..10 {
let job = state.get_job(&format!("concurrent_stop_{}", i)).unwrap();
assert_eq!(job.status, TrainingStatus::Stopped as i32);
}
}
/// TEST 9: Stop response message validation
#[tokio::test]
async fn test_stop_response_message() {
let state = MockTrainingState::default();
let job = test_fixtures::create_test_job(
"message_test",
"PPO",
TrainingStatus::Running,
);
state.add_job(job);
let addr = start_mock_server(state).await;
let channel = Channel::from_shared(format!("http://{}", addr))
.unwrap()
.connect()
.await
.unwrap();
let mut client = MlTrainingServiceClient::new(channel);
let request = StopTrainingRequest {
job_id: "message_test".to_string(),
reason: "Test".to_string(),
};
let response = client.stop_training(request).await.unwrap();
let result = response.into_inner();
assert!(result.success);
assert!(!result.message.is_empty());
assert!(result.message.contains("message_test"));
}
/// TEST 10: Stop job cleanup verification
#[tokio::test]
async fn test_stop_cleanup() {
let state = MockTrainingState::default();
let job = test_fixtures::create_test_job(
"cleanup_test",
"TFT",
TrainingStatus::Running,
);
state.add_job(job);
let addr = start_mock_server(state.clone()).await;
let channel = Channel::from_shared(format!("http://{}", addr))
.unwrap()
.connect()
.await
.unwrap();
let mut client = MlTrainingServiceClient::new(channel);
let request = StopTrainingRequest {
job_id: "cleanup_test".to_string(),
reason: "Cleanup test".to_string(),
};
let response = client.stop_training(request).await;
assert!(response.is_ok());
// Verify job still exists but is stopped
let job = state.get_job("cleanup_test");
assert!(job.is_some());
assert_eq!(job.unwrap().status, TrainingStatus::Stopped as i32);
}