system_status_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::make_request_span;
18
use crate::logging::TraceParent;
19
20
use crate::metrics::MetricsRegistry;
use crate::traits::DistributedRuntimeProvider;
21
use axum::{body, http::StatusCode, response::IntoResponse, routing::get, Router};
22
23
use serde_json::json;
use std::collections::HashMap;
24
use std::sync::Arc;
25
26
use std::sync::OnceLock;
use std::time::Instant;
27
use tokio::{net::TcpListener, task::JoinHandle};
28
use tokio_util::sync::CancellationToken;
29
30
use tower_http::trace::DefaultMakeSpan;
use tower_http::trace::TraceLayer;
31

32
/// System status server information containing socket address and handle
33
#[derive(Debug)]
34
pub struct SystemStatusServerInfo {
35
36
37
38
    pub socket_addr: std::net::SocketAddr,
    pub handle: Option<Arc<JoinHandle<()>>>,
}

39
impl SystemStatusServerInfo {
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
    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()
    }
}

60
impl Clone for SystemStatusServerInfo {
61
62
63
64
65
66
67
68
    fn clone(&self) -> Self {
        Self {
            socket_addr: self.socket_addr,
            handle: self.handle.clone(),
        }
    }
}

69
70
/// System status server state containing metrics and uptime tracking
pub struct SystemStatusState {
71
72
73
    // global drt registry is for printing out the entire Prometheus format output
    root_drt: Arc<crate::DistributedRuntime>,
    start_time: OnceLock<Instant>,
74
    uptime_gauge: prometheus::Gauge,
75
76
}

77
78
impl SystemStatusState {
    /// Create new system status server state with the provided metrics registry
79
    pub fn new(drt: Arc<crate::DistributedRuntime>) -> anyhow::Result<Self> {
80
81
        // 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
82
83
        let uptime_gauge = drt.as_ref().create_gauge(
            "dynamo_uptime_seconds",
84
85
86
87
88
89
90
91
92
93
            "Total uptime of the DistributedRuntime in seconds",
            &[],
        )?;
        let state = Self {
            root_drt: drt,
            start_time: OnceLock::new(),
            uptime_gauge,
        };
        Ok(state)
    }
94

95
96
97
98
99
100
    /// 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")
    }
101

102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
    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");
        }
122
123
124
    }
}

125
126
/// Start system status server with metrics support
pub async fn spawn_system_status_server(
127
128
129
130
    host: &str,
    port: u16,
    cancel_token: CancellationToken,
    drt: Arc<crate::DistributedRuntime>,
131
) -> anyhow::Result<(std::net::SocketAddr, tokio::task::JoinHandle<()>)> {
132
133
    // Create system status server state with the provided metrics registry
    let server_state = Arc::new(SystemStatusState::new(drt)?);
134
135
136
137
138
139
140
141
142
143
144
145
146
147
    let health_path = server_state
        .drt()
        .system_health
        .lock()
        .unwrap()
        .health_path
        .clone();
    let live_path = server_state
        .drt()
        .system_health
        .lock()
        .unwrap()
        .live_path
        .clone();
148

149
150
151
152
153
    // Initialize the start time
    server_state
        .initialize_start_time()
        .map_err(|e| anyhow::anyhow!("Failed to initialize start time: {}", e))?;

154
    let app = Router::new()
155
        .route(
156
            &health_path,
157
158
            get({
                let state = Arc::clone(&server_state);
159
                move || health_handler(state)
160
161
162
            }),
        )
        .route(
163
            &live_path,
164
165
            get({
                let state = Arc::clone(&server_state);
166
                move || health_handler(state)
167
168
            }),
        )
169
170
171
172
        .route(
            "/metrics",
            get({
                let state = Arc::clone(&server_state);
173
                move || metrics_handler(state)
174
            }),
175
        )
176
177
178
179
180
        .fallback(|| async {
            tracing::info!("[fallback handler] called");
            (StatusCode::NOT_FOUND, "Route not found").into_response()
        })
        .layer(TraceLayer::new_for_http().make_span_with(make_request_span));
181
182

    let address = format!("{}:{}", host, port);
183
    tracing::info!("[spawn_system_status_server] binding to: {}", address);
184
185
186
187
188

    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()?;
189
            tracing::info!(
190
                "[spawn_system_status_server] system status server bound to: {}",
191
192
193
                actual_address
            );
            (listener, actual_address)
194
195
196
197
198
199
        }
        Err(e) => {
            tracing::error!("Failed to bind to address {}: {}", address, e);
            return Err(anyhow::anyhow!("Failed to bind to address: {}", e));
        }
    };
200
    let (listener, actual_address) = listener;
201
202

    let observer = cancel_token.child_token();
203
204
205
206
207
208
    // 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
        {
209
            tracing::error!("System status server error: {}", e);
210
211
212
        }
    });
    Ok((actual_address, handle))
213
214
}

215
/// Health handler
216
217
#[tracing::instrument(skip_all, level = "trace")]
async fn health_handler(state: Arc<SystemStatusState>) -> impl IntoResponse {
218
219
220
221
222
223
    let (mut healthy, endpoints) = state
        .drt()
        .system_health
        .lock()
        .unwrap()
        .get_health_status();
224
225
    let uptime = match state.uptime() {
        Ok(uptime_state) => Some(uptime_state),
226
227
        Err(e) => {
            tracing::error!("Failed to get uptime: {}", e);
228
229
            healthy = false;
            None
230
        }
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
    };

    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())
249
}
250
251

/// Metrics handler with DistributedRuntime uptime
252
253
#[tracing::instrument(skip_all, level = "trace")]
async fn metrics_handler(state: Arc<SystemStatusState>) -> impl IntoResponse {
254
    // Update the uptime gauge with current value
255
    state.update_uptime_gauge();
256

257
258
259
    // Get metrics from the registry
    match state.drt().prometheus_metrics_fmt() {
        Ok(response) => (StatusCode::OK, response),
260
        Err(e) => {
261
            tracing::error!("Failed to get metrics from registry: {}", e);
262
263
            (
                StatusCode::INTERNAL_SERVER_ERROR,
264
                "Failed to get metrics".to_string(),
265
266
267
268
269
            )
        }
    }
}

270
271
// Regular tests: cargo test system_status_server --lib
// Integration tests: cargo test system_status_server --lib --features integration
272
273
274
275
276
277
278
279
280
281
282

#[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()
}

283
284
285
#[cfg(test)]
mod tests {
    use super::*;
286
    use crate::logging::tests::load_log;
287
    use crate::metrics::MetricsRegistry;
288
289
290
    use anyhow::{anyhow, Result};
    use chrono::{DateTime, Utc};
    use jsonschema::{Draft, JSONSchema};
291
    use rstest::rstest;
292
293
294
    use serde_json::Value;
    use std::fs::File;
    use std::io::{BufRead, BufReader};
295
    use std::sync::Arc;
296
    use stdio_override::*;
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
    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"
        );
    }

329
    #[cfg(feature = "integration")]
330
    #[tokio::test]
331
332
333
    async fn test_runtime_metrics_initialization_and_namespace() {
        // Test that metrics have correct namespace
        let drt = create_test_drt_async().await;
334
        let runtime_metrics = SystemStatusState::new(Arc::new(drt)).unwrap();
335

336
337
        // Initialize start time
        runtime_metrics.initialize_start_time().unwrap();
338

339
        runtime_metrics.uptime_gauge.set(42.0);
340

341
342
        let response = runtime_metrics.drt().prometheus_metrics_fmt().unwrap();
        println!("Full metrics response:\n{}", response);
343

344
        let expected = "\
345
346
347
# HELP dynamo_component_dynamo_uptime_seconds Total uptime of the DistributedRuntime in seconds
# TYPE dynamo_component_dynamo_uptime_seconds gauge
dynamo_component_dynamo_uptime_seconds 42
348
349
";
        assert_eq!(response, expected);
350
351
    }

352
    #[cfg(feature = "integration")]
353
    #[tokio::test]
354
355
356
    async fn test_start_time_initialization() {
        // Test that start time can only be initialized once
        let drt = create_test_drt_async().await;
357
        let runtime_metrics = SystemStatusState::new(Arc::new(drt)).unwrap();
358

359
360
        // First initialization should succeed
        assert!(runtime_metrics.initialize_start_time().is_ok());
361

362
363
        // Second initialization should fail
        assert!(runtime_metrics.initialize_start_time().is_err());
364

365
366
367
        // Uptime should work after initialization
        let _uptime = runtime_metrics.uptime().unwrap();
        // If we get here, uptime calculation works correctly
368
    }
369

370
    #[rstest]
371
372
373
374
375
376
377
378
379
380
381
    #[case("ready", 200, "ready", None, None, 3)]
    #[case("notready", 503, "notready", None, None, 3)]
    #[case("ready", 200, "ready", Some("/custom/health"), Some("/custom/live"), 5)]
    #[case(
        "notready",
        503,
        "notready",
        Some("/custom/health"),
        Some("/custom/live"),
        5
    )]
382
    #[tokio::test]
383
    #[cfg(feature = "integration")]
384
385
386
387
    async fn test_health_endpoints(
        #[case] starting_health_status: &'static str,
        #[case] expected_status: u16,
        #[case] expected_body: &'static str,
388
389
390
        #[case] custom_health_path: Option<&'static str>,
        #[case] custom_live_path: Option<&'static str>,
        #[case] expected_num_tests: usize,
391
392
393
394
395
396
397
398
399
    ) {
        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

400
401
        crate::logging::init();

402
403
        #[allow(clippy::redundant_closure_call)]
        temp_env::async_with_vars(
404
405
406
407
408
409
410
411
            [
                (
                    "DYN_SYSTEM_STARTING_HEALTH_STATUS",
                    Some(starting_health_status),
                ),
                ("DYN_SYSTEM_HEALTH_PATH", custom_health_path),
                ("DYN_SYSTEM_LIVE_PATH", custom_live_path),
            ],
412
413
414
415
416
417
418
419
            (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();
420
421
422
423
                let (addr, _) =
                    spawn_system_status_server("127.0.0.1", 0, cancel_token.clone(), drt)
                        .await
                        .unwrap();
424
425
426
427
                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();
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450

                // Prepare test cases
                let mut test_cases = vec![];
                if custom_health_path.is_none() {
                    // When using default paths, test the default paths
                    test_cases.push(("/health", expected_status, expected_body));
                } else {
                    // When using custom paths, default paths should not exist
                    test_cases.push(("/health", 404, "Route not found"));
                    test_cases.push((custom_health_path.unwrap(), expected_status, expected_body));
                }
                if custom_live_path.is_none() {
                    // When using default paths, test the default paths
                    test_cases.push(("/live", expected_status, expected_body));
                } else {
                    // When using custom paths, default paths should not exist
                    test_cases.push(("/live", 404, "Route not found"));
                    test_cases.push((custom_live_path.unwrap(), expected_status, expected_body));
                }
                test_cases.push(("/someRandomPathNotFoundHere", 404, "Route not found"));
                assert_eq!(test_cases.len(), expected_num_tests);

                for (path, expect_status, expect_body) in test_cases {
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
476
                    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
                    );
                    assert_eq!(
                        status, expect_status,
                        "Response: status={}, body={:?}",
                        status, body
                    );
                    assert!(
                        body.contains(expect_body),
                        "Response: status={}, body={:?}",
                        status,
                        body
                    );
                }
            })(),
        )
        .await;
    }

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
    #[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();
506
507
508
509
                let (addr, _) =
                    spawn_system_status_server("127.0.0.1", 0, cancel_token.clone(), drt)
                        .await
                        .unwrap();
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
537
538
                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(())
    }

539
540
541
542
543
    #[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;
544
        let runtime_metrics = SystemStatusState::new(Arc::new(drt)).unwrap();
545
546
547
548
549
550
551
552

        // 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")]
553
    #[tokio::test]
554
    async fn test_spawn_system_status_server_endpoints() {
555
        // use reqwest for HTTP requests
556
557
558
559
560
561
        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) =
562
                    spawn_system_status_server("127.0.0.1", 0, cancel_token.clone(), Arc::new(drt))
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
592
593
                        .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
                    );
594
                }
595
596
597
598
599
600
601
602
603
604
605
606
607
608
                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;
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
640
641

    #[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"
        );
    }
642
}