http_server.rs 23.2 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
// SPDX-FileCopyrightText: Copyright (c) 2024-2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
// SPDX-License-Identifier: Apache-2.0
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

16
use crate::config::HealthStatus;
17
use crate::logging::TraceParent;
18
19
use crate::metrics::MetricsRegistry;
use crate::traits::DistributedRuntimeProvider;
20
use axum::{body, http::StatusCode, response::IntoResponse, routing::get, Router};
21
22
use serde_json::json;
use std::collections::HashMap;
23
use std::sync::Arc;
24
25
use std::sync::OnceLock;
use std::time::Instant;
26
use tokio::{net::TcpListener, task::JoinHandle};
27
28
use tokio_util::sync::CancellationToken;
use tracing;
29
use tracing::Instrument;
30

31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
/// HTTP server information containing socket address and handle
#[derive(Debug)]
pub struct HttpServerInfo {
    pub socket_addr: std::net::SocketAddr,
    pub handle: Option<Arc<JoinHandle<()>>>,
}

impl HttpServerInfo {
    pub fn new(socket_addr: std::net::SocketAddr, handle: Option<JoinHandle<()>>) -> Self {
        Self {
            socket_addr,
            handle: handle.map(Arc::new),
        }
    }

    pub fn address(&self) -> String {
        self.socket_addr.to_string()
    }

    pub fn hostname(&self) -> String {
        self.socket_addr.ip().to_string()
    }

    pub fn port(&self) -> u16 {
        self.socket_addr.port()
    }
}

impl Clone for HttpServerInfo {
    fn clone(&self) -> Self {
        Self {
            socket_addr: self.socket_addr,
            handle: self.handle.clone(),
        }
    }
}

68
69
pub struct HttpMetricsRegistry {
    pub drt: Arc<crate::DistributedRuntime>,
70
71
}

72
73
74
75
76
impl crate::traits::DistributedRuntimeProvider for HttpMetricsRegistry {
    fn drt(&self) -> &crate::DistributedRuntime {
        &self.drt
    }
}
77

78
79
impl MetricsRegistry for HttpMetricsRegistry {
    fn basename(&self) -> String {
80
        "dynamo".to_string()
81
82
    }

83
84
    fn parent_hierarchy(&self) -> Vec<String> {
        [self.drt().parent_hierarchy(), vec![self.drt().basename()]].concat()
85
86
87
    }
}

88
/// HTTP server state containing metrics and uptime tracking
89
pub struct HttpServerState {
90
91
92
93
    // global drt registry is for printing out the entire Prometheus format output
    root_drt: Arc<crate::DistributedRuntime>,
    start_time: OnceLock<Instant>,
    uptime_gauge: Arc<prometheus::Gauge>,
94
95
96
}

impl HttpServerState {
97
    /// Create new HTTP server state with the provided metrics registry
98
    pub fn new(drt: Arc<crate::DistributedRuntime>) -> anyhow::Result<Self> {
99
        let http_metrics_registry = Arc::new(HttpMetricsRegistry { drt: drt.clone() });
100
101
        // Note: This metric is created at the DRT level (no namespace), so we manually add "dynamo_" prefix
        // to maintain consistency with the project's metric naming convention
102
        let uptime_gauge = http_metrics_registry.as_ref().create_gauge(
103
            "system_uptime_seconds",
104
105
106
107
108
109
110
111
112
113
            "Total uptime of the DistributedRuntime in seconds",
            &[],
        )?;
        let state = Self {
            root_drt: drt,
            start_time: OnceLock::new(),
            uptime_gauge,
        };
        Ok(state)
    }
114

115
116
117
118
119
120
    /// Initialize the start time (can only be called once)
    pub fn initialize_start_time(&self) -> Result<(), &'static str> {
        self.start_time
            .set(Instant::now())
            .map_err(|_| "Start time already initialized")
    }
121

122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
    pub fn uptime(&self) -> Result<std::time::Duration, &'static str> {
        self.start_time
            .get()
            .ok_or("Start time not initialized")
            .map(|start_time| start_time.elapsed())
    }

    /// Get a reference to the distributed runtime
    pub fn drt(&self) -> &crate::DistributedRuntime {
        &self.root_drt
    }

    /// Update the uptime gauge with current value
    pub fn update_uptime_gauge(&self) {
        if let Ok(uptime) = self.uptime() {
            let uptime_seconds = uptime.as_secs_f64();
            self.uptime_gauge.set(uptime_seconds);
        } else {
            tracing::warn!("Failed to update uptime gauge: start time not initialized");
        }
142
143
144
    }
}

145
/// Start HTTP server with metrics support
146
pub async fn spawn_http_server(
147
148
149
150
    host: &str,
    port: u16,
    cancel_token: CancellationToken,
    drt: Arc<crate::DistributedRuntime>,
151
) -> anyhow::Result<(std::net::SocketAddr, tokio::task::JoinHandle<()>)> {
152
    // Create HTTP server state with the provided metrics registry
153
154
    let server_state = Arc::new(HttpServerState::new(drt)?);

155
156
157
158
159
    // Initialize the start time
    server_state
        .initialize_start_time()
        .map_err(|e| anyhow::anyhow!("Failed to initialize start time: {}", e))?;

160
    let app = Router::new()
161
162
163
164
        .route(
            "/health",
            get({
                let state = Arc::clone(&server_state);
165
                move |tracing_ctx| health_handler(state, "health", tracing_ctx)
166
167
168
169
170
171
            }),
        )
        .route(
            "/live",
            get({
                let state = Arc::clone(&server_state);
172
                move |tracing_ctx| health_handler(state, "live", tracing_ctx)
173
174
            }),
        )
175
176
177
178
        .route(
            "/metrics",
            get({
                let state = Arc::clone(&server_state);
179
                move |tracing_ctx| metrics_handler(state, "metrics", tracing_ctx)
180
            }),
181
        )
182
183
184
185
186
187
188
189
190
191
192
193
        .fallback(|tracing_ctx: TraceParent| {
            async {
                tracing::info!("[fallback handler] called");
                (StatusCode::NOT_FOUND, "Route not found").into_response()
            }
            .instrument(tracing::trace_span!(
                "fallback handler",
                trace_id = tracing_ctx.trace_id,
                parent_id = tracing_ctx.parent_id,
                x_request_id = tracing_ctx.x_request_id,
                tracestate = tracing_ctx.tracestate
            ))
194
        });
195
196

    let address = format!("{}:{}", host, port);
197
    tracing::info!("[spawn_http_server] binding to: {}", address);
198
199
200
201
202

    let listener = match TcpListener::bind(&address).await {
        Ok(listener) => {
            // get the actual address and port, print in debug level
            let actual_address = listener.local_addr()?;
203
204
205
206
207
            tracing::info!(
                "[spawn_http_server] HTTP server bound to: {}",
                actual_address
            );
            (listener, actual_address)
208
209
210
211
212
213
        }
        Err(e) => {
            tracing::error!("Failed to bind to address {}: {}", address, e);
            return Err(anyhow::anyhow!("Failed to bind to address: {}", e));
        }
    };
214
    let (listener, actual_address) = listener;
215
216

    let observer = cancel_token.child_token();
217
218
219
220
221
222
223
224
225
226
    // Spawn the server in the background and return the handle
    let handle = tokio::spawn(async move {
        if let Err(e) = axum::serve(listener, app)
            .with_graceful_shutdown(observer.cancelled_owned())
            .await
        {
            tracing::error!("HTTP server error: {}", e);
        }
    });
    Ok((actual_address, handle))
227
228
}

229
/// Health handler
230
231
232
233
234
235
236
237
238
239
#[tracing::instrument(skip_all, level="trace", fields(route= %route,
						      trace_id = trace_parent.trace_id,
						      parent_id = trace_parent.parent_id,
						      x_request_id= trace_parent.x_request_id,
						      tracestate= trace_parent.tracestate))]
async fn health_handler(
    state: Arc<HttpServerState>,
    route: &'static str,       // Used for tracing only
    trace_parent: TraceParent, // Used for tracing only
) -> impl IntoResponse {
240
241
242
243
    let system_health = state.drt().system_health.lock().await;
    let (mut healthy, endpoints) = system_health.get_health_status();
    let uptime = match state.uptime() {
        Ok(uptime_state) => Some(uptime_state),
244
245
        Err(e) => {
            tracing::error!("Failed to get uptime: {}", e);
246
247
            healthy = false;
            None
248
        }
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
    };

    let healthy_string = if healthy { "ready" } else { "notready" };
    let status_code = if healthy {
        StatusCode::OK
    } else {
        StatusCode::SERVICE_UNAVAILABLE
    };

    let response = json!({
        "status": healthy_string,
        "uptime": uptime,
        "endpoints": endpoints
    });

    tracing::trace!("Response {}", response.to_string());

    (status_code, response.to_string())
267
}
268
269

/// Metrics handler with DistributedRuntime uptime
270
271
272
273
274
275
276
277
278
279
#[tracing::instrument(skip_all, level="trace", fields(route= %route,
						      trace_id = trace_parent.trace_id,
						      parent_id = trace_parent.parent_id,
						      x_request_id = trace_parent.x_request_id,
                                                      tracestate = trace_parent.tracestate))]
async fn metrics_handler(
    state: Arc<HttpServerState>,
    route: &'static str,       // Used for tracing only
    trace_parent: TraceParent, // Used for tracing only
) -> impl IntoResponse {
280
    // Update the uptime gauge with current value
281
    state.update_uptime_gauge();
282

283
284
285
    // Get metrics from the registry
    match state.drt().prometheus_metrics_fmt() {
        Ok(response) => (StatusCode::OK, response),
286
        Err(e) => {
287
            tracing::error!("Failed to get metrics from registry: {}", e);
288
289
            (
                StatusCode::INTERNAL_SERVER_ERROR,
290
                "Failed to get metrics".to_string(),
291
292
293
294
295
            )
        }
    }
}

296
297
298
299
300
301
302
303
304
305
306
307
308
// Regular tests: cargo test http_server --lib
// Integration tests: cargo test http_server --lib --features integration

#[cfg(test)]
/// Helper function to create a DRT instance for async testing
/// Uses the test-friendly constructor without discovery
async fn create_test_drt_async() -> crate::DistributedRuntime {
    let rt = crate::Runtime::from_current().unwrap();
    crate::DistributedRuntime::from_settings_without_discovery(rt)
        .await
        .unwrap()
}

309
310
311
#[cfg(test)]
mod tests {
    use super::*;
312
    use crate::logging::tests::load_log;
313
    use crate::metrics::MetricsRegistry;
314
315
316
    use anyhow::{anyhow, Result};
    use chrono::{DateTime, Utc};
    use jsonschema::{Draft, JSONSchema};
317
    use rstest::rstest;
318
319
320
    use serde_json::Value;
    use std::fs::File;
    use std::io::{BufRead, BufReader};
321
    use std::sync::Arc;
322
    use stdio_override::*;
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
    use tokio::time::{sleep, Duration};

    #[tokio::test]
    async fn test_http_server_lifecycle() {
        let cancel_token = CancellationToken::new();
        let cancel_token_for_server = cancel_token.clone();

        // Test basic HTTP server lifecycle without DistributedRuntime
        let app = Router::new().route("/test", get(|| async { (StatusCode::OK, "test") }));

        // start HTTP server
        let server_handle = tokio::spawn(async move {
            let listener = TcpListener::bind("127.0.0.1:0").await.unwrap();
            let _ = axum::serve(listener, app)
                .with_graceful_shutdown(cancel_token_for_server.cancelled_owned())
                .await;
        });

        // wait for a while to let the server start
        sleep(Duration::from_millis(100)).await;

        // cancel token
        cancel_token.cancel();

        // wait for the server to shut down
        let result = tokio::time::timeout(Duration::from_secs(5), server_handle).await;
        assert!(
            result.is_ok(),
            "HTTP server should shut down when cancel token is cancelled"
        );
    }

355
    #[cfg(feature = "integration")]
356
    #[tokio::test]
357
358
359
360
    async fn test_runtime_metrics_initialization_and_namespace() {
        // Test that metrics have correct namespace
        let drt = create_test_drt_async().await;
        let runtime_metrics = HttpServerState::new(Arc::new(drt)).unwrap();
361

362
363
        // Initialize start time
        runtime_metrics.initialize_start_time().unwrap();
364

365
        runtime_metrics.uptime_gauge.set(42.0);
366

367
368
        let response = runtime_metrics.drt().prometheus_metrics_fmt().unwrap();
        println!("Full metrics response:\n{}", response);
369

370
        let expected = "\
371
372
373
# HELP dynamo_system_uptime_seconds Total uptime of the DistributedRuntime in seconds
# TYPE dynamo_system_uptime_seconds gauge
dynamo_system_uptime_seconds{namespace=\"dynamo\"} 42
374
375
";
        assert_eq!(response, expected);
376
377
    }

378
    #[cfg(feature = "integration")]
379
    #[tokio::test]
380
381
382
383
    async fn test_start_time_initialization() {
        // Test that start time can only be initialized once
        let drt = create_test_drt_async().await;
        let runtime_metrics = HttpServerState::new(Arc::new(drt)).unwrap();
384

385
386
        // First initialization should succeed
        assert!(runtime_metrics.initialize_start_time().is_ok());
387

388
389
        // Second initialization should fail
        assert!(runtime_metrics.initialize_start_time().is_err());
390

391
392
393
        // Uptime should work after initialization
        let _uptime = runtime_metrics.uptime().unwrap();
        // If we get here, uptime calculation works correctly
394
    }
395

396
397
398
399
    #[rstest]
    #[case("ready", 200, "ready")]
    #[case("notready", 503, "notready")]
    #[tokio::test]
400
    #[cfg(feature = "integration")]
401
402
403
404
405
406
407
408
409
410
411
412
413
    async fn test_health_endpoints(
        #[case] starting_health_status: &'static str,
        #[case] expected_status: u16,
        #[case] expected_body: &'static str,
    ) {
        use std::sync::Arc;
        use tokio::time::sleep;
        use tokio_util::sync::CancellationToken;
        // use tokio::io::{AsyncReadExt, AsyncWriteExt};
        // use reqwest for HTTP requests

        // Closure call is needed here to satisfy async_with_vars

414
415
        crate::logging::init();

416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
        #[allow(clippy::redundant_closure_call)]
        temp_env::async_with_vars(
            [(
                "DYN_SYSTEM_STARTING_HEALTH_STATUS",
                Some(starting_health_status),
            )],
            (async || {
                let runtime = crate::Runtime::from_settings().unwrap();
                let drt = Arc::new(
                    crate::DistributedRuntime::from_settings_without_discovery(runtime)
                        .await
                        .unwrap(),
                );
                let cancel_token = CancellationToken::new();
                let (addr, _) = spawn_http_server("127.0.0.1", 0, cancel_token.clone(), drt)
                    .await
                    .unwrap();
                println!("[test] Waiting for server to start...");
                sleep(std::time::Duration::from_millis(1000)).await;
                println!("[test] Server should be up, starting requests...");
                let client = reqwest::Client::new();
                for (path, expect_status, expect_body) in [
                    ("/health", expected_status, expected_body),
                    ("/live", expected_status, expected_body),
                    ("/someRandomPathNotFoundHere", 404, "Route not found"),
                ] {
                    println!("[test] Sending request to {}", path);
443
444
445
446
447
448
449
450
                    let traceparent_value =
                        "00-4bf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7-01";
                    let tracestate_value = "vendor1=opaqueValue1,vendor2=opaqueValue2";
                    let mut headers = reqwest::header::HeaderMap::new();
                    headers.insert(
                        reqwest::header::HeaderName.from_static("traceparent"),
                        reqwest::header::HeaderValue.from_str(traceparent_value)?,
                    );
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
                    let url = format!("http://{}{}", addr, path);
                    let response = client.get(&url).send().await.unwrap();
                    let status = response.status();
                    let body = response.text().await.unwrap();
                    println!(
                        "[test] Response for {}: status={}, body={:?}",
                        path, status, body
                    );
                    assert_eq!(
                        status, expect_status,
                        "Response: status={}, body={:?}",
                        status, body
                    );
                    assert!(
                        body.contains(expect_body),
                        "Response: status={}, body={:?}",
                        status,
                        body
                    );
                }
            })(),
        )
        .await;
    }

476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
    #[tokio::test]
    #[cfg(feature = "integration")]
    async fn test_health_endpoint_tracing() -> Result<()> {
        use std::sync::Arc;
        use tokio::time::sleep;
        use tokio_util::sync::CancellationToken;

        // Closure call is needed here to satisfy async_with_vars

        #[allow(clippy::redundant_closure_call)]
        let _ = temp_env::async_with_vars(
            [
                ("DYN_SYSTEM_STARTING_HEALTH_STATUS", Some("ready")),
                ("DYN_LOGGING_JSONL", Some("1")),
                ("DYN_LOG", Some("trace")),
            ],
            (async || {
                // TODO Add proper testing for
                // trace id and parent id

                crate::logging::init();

                let runtime = crate::Runtime::from_settings().unwrap();
                let drt = Arc::new(
                    crate::DistributedRuntime::from_settings_without_discovery(runtime)
                        .await
                        .unwrap(),
                );
                let cancel_token = CancellationToken::new();
                let (addr, _) = spawn_http_server("127.0.0.1", 0, cancel_token.clone(), drt)
                    .await
                    .unwrap();
                sleep(std::time::Duration::from_millis(1000)).await;
                let client = reqwest::Client::new();
                for path in [("/health"), ("/live"), ("/someRandomPathNotFoundHere")] {
                    let traceparent_value =
                        "00-4bf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7-01";
                    let tracestate_value = "vendor1=opaqueValue1,vendor2=opaqueValue2";
                    let mut headers = reqwest::header::HeaderMap::new();
                    headers.insert(
                        reqwest::header::HeaderName::from_static("traceparent"),
                        reqwest::header::HeaderValue::from_str(traceparent_value)?,
                    );
                    headers.insert(
                        reqwest::header::HeaderName::from_static("tracestate"),
                        reqwest::header::HeaderValue::from_str(tracestate_value)?,
                    );
                    let url = format!("http://{}{}", addr, path);
                    let response = client.get(&url).headers(headers).send().await.unwrap();
                    let status = response.status();
                    let body = response.text().await.unwrap();
                    tracing::info!(body = body, status = status.to_string());
                }

                Ok::<(), anyhow::Error>(())
            })(),
        )
        .await;
        Ok(())
    }

537
538
539
540
541
542
543
544
545
546
547
548
549
550
    #[cfg(feature = "integration")]
    #[tokio::test]
    async fn test_uptime_without_initialization() {
        // Test that uptime returns an error if start time is not initialized
        let drt = create_test_drt_async().await;
        let runtime_metrics = HttpServerState::new(Arc::new(drt)).unwrap();

        // This should return an error because start time is not initialized
        let result = runtime_metrics.uptime();
        assert!(result.is_err());
        assert_eq!(result.unwrap_err(), "Start time not initialized");
    }

    #[cfg(feature = "integration")]
551
552
553
    #[tokio::test]
    async fn test_spawn_http_server_endpoints() {
        // use reqwest for HTTP requests
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
        temp_env::async_with_vars(
            [("DYN_SYSTEM_STARTING_HEALTH_STATUS", Some("ready"))],
            async {
                let cancel_token = CancellationToken::new();
                let drt = create_test_drt_async().await;
                let (addr, server_handle) =
                    spawn_http_server("127.0.0.1", 0, cancel_token.clone(), Arc::new(drt))
                        .await
                        .unwrap();
                println!("[test] Waiting for server to start...");
                sleep(std::time::Duration::from_millis(1000)).await;
                println!("[test] Server should be up, starting requests...");
                let client = reqwest::Client::new();
                for (path, expect_200, expect_body) in [
                    ("/health", true, "ready"),
                    ("/live", true, "ready"),
                    ("/someRandomPathNotFoundHere", false, "Route not found"),
                ] {
                    println!("[test] Sending request to {}", path);
                    let url = format!("http://{}{}", addr, path);
                    let response = client.get(&url).send().await.unwrap();
                    let status = response.status();
                    let body = response.text().await.unwrap();
                    println!(
                        "[test] Response for {}: status={}, body={:?}",
                        path, status, body
                    );
                    if expect_200 {
                        assert_eq!(status, 200, "Response: status={}, body={:?}", status, body);
                    } else {
                        assert_eq!(status, 404, "Response: status={}, body={:?}", status, body);
                    }
                    assert!(
                        body.contains(expect_body),
                        "Response: status={}, body={:?}",
                        status,
                        body
                    );
592
                }
593
594
595
596
597
598
599
600
601
602
603
604
605
606
                cancel_token.cancel();
                match server_handle.await {
                    Ok(_) => println!("[test] Server shut down normally"),
                    Err(e) => {
                        if e.is_panic() {
                            println!("[test] Server panicked: {:?}", e);
                        } else {
                            println!("[test] Server cancelled: {:?}", e);
                        }
                    }
                }
            },
        )
        .await;
607
    }
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639

    #[cfg(feature = "integration")]
    #[tokio::test]
    async fn test_http_server_basic_functionality() {
        // Test basic HTTP server functionality without requiring etcd
        let cancel_token = CancellationToken::new();
        let cancel_token_for_server = cancel_token.clone();

        // Test basic HTTP server lifecycle
        let app = Router::new().route("/test", get(|| async { (StatusCode::OK, "test") }));

        // start HTTP server
        let server_handle = tokio::spawn(async move {
            let listener = TcpListener::bind("127.0.0.1:0").await.unwrap();
            let _ = axum::serve(listener, app)
                .with_graceful_shutdown(cancel_token_for_server.cancelled_owned())
                .await;
        });

        // wait for a while to let the server start
        sleep(Duration::from_millis(100)).await;

        // cancel token
        cancel_token.cancel();

        // wait for the server to shut down
        let result = tokio::time::timeout(Duration::from_secs(5), server_handle).await;
        assert!(
            result.is_ok(),
            "HTTP server should shut down when cancel token is cancelled"
        );
    }
640
}