logging.rs 58 KB
Newer Older
1
// SPDX-FileCopyrightText: Copyright (c) 2024-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
2
3
// SPDX-License-Identifier: Apache-2.0

Neelay Shah's avatar
Neelay Shah committed
4
//! Dynamo Distributed Logging Module.
5
6
7
//!
//! - Configuration loaded from:
//!   1. Environment variables (highest priority).
8
//!   2. Optional TOML file pointed to by the `DYN_LOGGING_CONFIG_PATH` environment variable.
Neelay Shah's avatar
Neelay Shah committed
9
//!   3. `/opt/dynamo/etc/logging.toml`.
10
11
//!
//! Logging can take two forms: `READABLE` or `JSONL`. The default is `READABLE`. `JSONL`
12
//! can be enabled by setting the `DYN_LOGGING_JSONL` environment variable to `1`.
13
//!
Ryan Olson's avatar
Ryan Olson committed
14
15
//! To use local timezone for logging timestamps, set the `DYN_LOG_USE_LOCAL_TZ` environment variable to `1`.
//!
16
//! Filters can be configured using the `DYN_LOG` environment variable or by setting the `filters`
17
//! key in the TOML configuration file. Filters are comma-separated key-value pairs where the key
18
//! is the crate or module name and the value is the log level. The default log level is `info`.
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
//!
//! Example:
//! ```toml
//! log_level = "error"
//!
//! [log_filters]
//! "test_logging" = "info"
//! "test_logging::api" = "trace"
//! ```

use std::collections::{BTreeMap, HashMap};
use std::sync::Once;

use figment::{
    Figment,
34
    providers::{Format, Serialized, Toml},
35
36
};
use serde::{Deserialize, Serialize};
37
use tracing::level_filters::LevelFilter;
38
use tracing::{Event, Subscriber};
39
use tracing_subscriber::EnvFilter;
Ryan Olson's avatar
Ryan Olson committed
40
41
42
43
use tracing_subscriber::fmt::time::FormatTime;
use tracing_subscriber::fmt::time::LocalTime;
use tracing_subscriber::fmt::time::SystemTime;
use tracing_subscriber::fmt::time::UtcTime;
44
use tracing_subscriber::fmt::{FmtContext, FormatFields};
45
use tracing_subscriber::fmt::{FormattedFields, format::Writer};
46
47
48
49
use tracing_subscriber::prelude::*;
use tracing_subscriber::registry::LookupSpan;
use tracing_subscriber::{filter::Directive, fmt};

50
use crate::config::{disable_ansi_logging, jsonl_logging_enabled};
51
use async_nats::{HeaderMap, HeaderValue};
52
use axum::extract::FromRequestParts;
53
54
use axum::http;
use axum::http::Request;
55
use axum::http::request::Parts;
56
57
58
use serde_json::Value;
use std::convert::Infallible;
use std::time::Instant;
59
use tower_http::trace::{DefaultMakeSpan, TraceLayer};
60
61
use tracing::Id;
use tracing::Span;
62
63
64
65
use tracing::field::Field;
use tracing::span;
use tracing_subscriber::Layer;
use tracing_subscriber::Registry;
66
67
68
69
70
use tracing_subscriber::field::Visit;
use tracing_subscriber::fmt::format::FmtSpan;
use tracing_subscriber::layer::Context;
use tracing_subscriber::registry::SpanData;
use uuid::Uuid;
71

72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
use opentelemetry::propagation::{Extractor, Injector, TextMapPropagator};
use opentelemetry::trace::TraceContextExt;
use opentelemetry::{global, trace::Tracer};
use opentelemetry_otlp::WithExportConfig;

use opentelemetry::trace::TracerProvider as _;
use opentelemetry::{Key, KeyValue};
use opentelemetry_sdk::Resource;
use opentelemetry_sdk::trace::SdkTracerProvider;
use tracing::error;
use tracing_subscriber::layer::SubscriberExt;
// use tracing_subscriber::Registry;

use std::time::Duration;
use tracing::{info, instrument};
use tracing_opentelemetry::OpenTelemetrySpanExt;
use tracing_subscriber::util::SubscriberInitExt;

90
use crate::config::environment_names::logging as env_logging;
91

92
93
use dynamo_config::env_is_truthy;

94
/// Default log level
95
const DEFAULT_FILTER_LEVEL: &str = "info";
96

97
98
99
100
101
102
/// Default OTLP endpoint
const DEFAULT_OTLP_ENDPOINT: &str = "http://localhost:4317";

/// Default service name
const DEFAULT_OTEL_SERVICE_NAME: &str = "dynamo";

103
104
105
106
107
108
109
110
111
112
113
114
/// Once instance to ensure the logger is only initialized once
static INIT: Once = Once::new();

#[derive(Serialize, Deserialize, Debug)]
struct LoggingConfig {
    log_level: String,
    log_filters: HashMap<String, String>,
}
impl Default for LoggingConfig {
    fn default() -> Self {
        LoggingConfig {
            log_level: DEFAULT_FILTER_LEVEL.to_string(),
Ryan Olson's avatar
Ryan Olson committed
115
116
117
118
119
120
            log_filters: HashMap::from([
                ("h2".to_string(), "error".to_string()),
                ("tower".to_string(), "error".to_string()),
                ("hyper_util".to_string(), "error".to_string()),
                ("neli".to_string(), "error".to_string()),
                ("async_nats".to_string(), "error".to_string()),
121
122
123
124
125
126
                ("rustls".to_string(), "error".to_string()),
                ("tokenizers".to_string(), "error".to_string()),
                ("axum".to_string(), "error".to_string()),
                ("tonic".to_string(), "error".to_string()),
                ("mistralrs_core".to_string(), "error".to_string()),
                ("hf_hub".to_string(), "error".to_string()),
127
128
129
                ("opentelemetry".to_string(), "error".to_string()),
                ("opentelemetry-otlp".to_string(), "error".to_string()),
                ("opentelemetry_sdk".to_string(), "error".to_string()),
Ryan Olson's avatar
Ryan Olson committed
130
            ]),
131
132
133
134
        }
    }
}

135
/// Check if OTLP trace exporting is enabled (accepts: "1", "true", "on", "yes" - case insensitive)
136
fn otlp_exporter_enabled() -> bool {
137
    env_is_truthy(env_logging::otlp::OTEL_EXPORT_ENABLED)
138
139
}

140
141
/// Get the service name from environment or use default
fn get_service_name() -> String {
142
143
    std::env::var(env_logging::otlp::OTEL_SERVICE_NAME)
        .unwrap_or_else(|_| DEFAULT_OTEL_SERVICE_NAME.to_string())
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
}

/// Validate a given trace ID according to W3C Trace Context specifications.
/// A valid trace ID is a 32-character hexadecimal string (lowercase).
pub fn is_valid_trace_id(trace_id: &str) -> bool {
    trace_id.len() == 32 && trace_id.chars().all(|c| c.is_ascii_hexdigit())
}

/// Validate a given span ID according to W3C Trace Context specifications.
/// A valid span ID is a 16-character hexadecimal string (lowercase).
pub fn is_valid_span_id(span_id: &str) -> bool {
    span_id.len() == 16 && span_id.chars().all(|c| c.is_ascii_hexdigit())
}

pub struct DistributedTraceIdLayer;

160
#[derive(Debug, Clone, Serialize, Deserialize)]
161
pub struct DistributedTraceContext {
162
163
164
165
166
167
168
169
170
    pub trace_id: String,
    pub span_id: String,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub parent_id: Option<String>,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub tracestate: Option<String>,
    #[serde(skip)]
    start: Option<Instant>,
    #[serde(skip)]
171
    end: Option<Instant>,
172
173
174
175
176
177
    #[serde(skip_serializing_if = "Option::is_none")]
    pub x_request_id: Option<String>,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub x_dynamo_request_id: Option<String>,
}

178
179
180
181
182
183
184
185
186
187
188
/// Pending context data collected in on_new_span, to be finalized in on_enter
#[derive(Debug, Clone)]
struct PendingDistributedTraceContext {
    trace_id: Option<String>,
    span_id: Option<String>,
    parent_id: Option<String>,
    tracestate: Option<String>,
    x_request_id: Option<String>,
    x_dynamo_request_id: Option<String>,
}

189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
impl DistributedTraceContext {
    /// Create a traceparent string from the context
    pub fn create_traceparent(&self) -> String {
        format!("00-{}-{}-01", self.trace_id, self.span_id)
    }
}

/// Parse a traceparent string into its components
pub fn parse_traceparent(traceparent: &str) -> (Option<String>, Option<String>) {
    let pieces: Vec<_> = traceparent.split('-').collect();
    if pieces.len() != 4 {
        return (None, None);
    }
    let trace_id = pieces[1];
    let parent_id = pieces[2];

    if !is_valid_trace_id(trace_id) || !is_valid_span_id(parent_id) {
        return (None, None);
    }

    (Some(trace_id.to_string()), Some(parent_id.to_string()))
210
211
}

212
#[derive(Debug, Clone, Default)]
213
214
215
216
217
pub struct TraceParent {
    pub trace_id: Option<String>,
    pub parent_id: Option<String>,
    pub tracestate: Option<String>,
    pub x_request_id: Option<String>,
218
    pub x_dynamo_request_id: Option<String>,
219
220
}

221
222
223
224
225
226
227
228
229
pub trait GenericHeaders {
    fn get(&self, key: &str) -> Option<&str>;
}

impl GenericHeaders for async_nats::HeaderMap {
    fn get(&self, key: &str) -> Option<&str> {
        async_nats::HeaderMap::get(self, key).map(|value| value.as_str())
    }
}
230

231
232
233
234
235
236
237
238
impl GenericHeaders for http::HeaderMap {
    fn get(&self, key: &str) -> Option<&str> {
        http::HeaderMap::get(self, key).and_then(|value| value.to_str().ok())
    }
}

impl TraceParent {
    pub fn from_headers<H: GenericHeaders>(headers: &H) -> TraceParent {
239
240
241
        let mut trace_id = None;
        let mut parent_id = None;
        let mut tracestate = None;
242
243
244
245
246
        let mut x_request_id = None;
        let mut x_dynamo_request_id = None;

        if let Some(header_value) = headers.get("traceparent") {
            (trace_id, parent_id) = parse_traceparent(header_value);
247
248
        }

249
250
        if let Some(header_value) = headers.get("x-request-id") {
            x_request_id = Some(header_value.to_string());
251
252
        }

253
254
255
256
257
258
259
        if let Some(header_value) = headers.get("tracestate") {
            tracestate = Some(header_value.to_string());
        }

        if let Some(header_value) = headers.get("x-dynamo-request-id") {
            x_dynamo_request_id = Some(header_value.to_string());
        }
260

261
262
263
264
        // Validate UUID format
        let x_dynamo_request_id =
            x_dynamo_request_id.filter(|id| uuid::Uuid::parse_str(id).is_ok());
        TraceParent {
265
266
267
268
            trace_id,
            parent_id,
            tracestate,
            x_request_id,
269
270
            x_dynamo_request_id,
        }
271
272
273
    }
}

274
275
276
277
278
279
280
// Takes Axum request and returning a span
pub fn make_request_span<B>(req: &Request<B>) -> Span {
    let method = req.method();
    let uri = req.uri();
    let version = format!("{:?}", req.version());
    let trace_parent = TraceParent::from_headers(req.headers());

281
282
    let otel_context = extract_otel_context_from_http_headers(req.headers());

283
    let span = tracing::info_span!(
284
285
286
287
288
289
290
        "http-request",
        method = %method,
        uri = %uri,
        version = %version,
        trace_id = trace_parent.trace_id,
        parent_id = trace_parent.parent_id,
        x_request_id = trace_parent.x_request_id,
291
        x_dynamo_request_id = trace_parent.x_dynamo_request_id,
292
293
    );

294
295
296
297
    if let Some(context) = otel_context {
        let _ = span.set_parent(context);
    }

298
299
300
    span
}

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
329
330
331
332
333
334
335
336
/// Extract OpenTelemetry context from HTTP headers for distributed tracing
fn extract_otel_context_from_http_headers(
    headers: &http::HeaderMap,
) -> Option<opentelemetry::Context> {
    let traceparent_value = headers.get("traceparent")?.to_str().ok()?;

    struct HttpHeaderExtractor<'a>(&'a http::HeaderMap);

    impl<'a> Extractor for HttpHeaderExtractor<'a> {
        fn get(&self, key: &str) -> Option<&str> {
            self.0.get(key).and_then(|v| v.to_str().ok())
        }

        fn keys(&self) -> Vec<&str> {
            vec!["traceparent", "tracestate"]
                .into_iter()
                .filter(|&key| self.0.get(key).is_some())
                .collect()
        }
    }

    // Early return if traceparent is empty
    if traceparent_value.is_empty() {
        return None;
    }

    let extractor = HttpHeaderExtractor(headers);
    let otel_context = TRACE_PROPAGATOR.extract(&extractor);

    if otel_context.span().span_context().is_valid() {
        Some(otel_context)
    } else {
        None
    }
}

337
338
339
340
341
342
/// Create a handle_payload span from NATS headers with component context
pub fn make_handle_payload_span(
    headers: &async_nats::HeaderMap,
    component: &str,
    endpoint: &str,
    namespace: &str,
343
    instance_id: u64,
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
) -> Span {
    let (otel_context, trace_id, parent_span_id) = extract_otel_context_from_nats_headers(headers);
    let trace_parent = TraceParent::from_headers(headers);

    if let (Some(trace_id), Some(parent_id)) = (trace_id.as_ref(), parent_span_id.as_ref()) {
        let span = tracing::info_span!(
            "handle_payload",
            trace_id = trace_id.as_str(),
            parent_id = parent_id.as_str(),
            x_request_id = trace_parent.x_request_id,
            x_dynamo_request_id = trace_parent.x_dynamo_request_id,
            tracestate = trace_parent.tracestate,
            component = component,
            endpoint = endpoint,
            namespace = namespace,
            instance_id = instance_id,
        );
361

362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
        if let Some(context) = otel_context {
            let _ = span.set_parent(context);
        }
        span
    } else {
        tracing::info_span!(
            "handle_payload",
            x_request_id = trace_parent.x_request_id,
            x_dynamo_request_id = trace_parent.x_dynamo_request_id,
            tracestate = trace_parent.tracestate,
            component = component,
            endpoint = endpoint,
            namespace = namespace,
            instance_id = instance_id,
        )
    }
}

380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
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
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
/// Create a handle_payload span from TCP/HashMap headers with component context
pub fn make_handle_payload_span_from_tcp_headers(
    headers: &std::collections::HashMap<String, String>,
    component: &str,
    endpoint: &str,
    namespace: &str,
    instance_id: u64,
) -> Span {
    let (otel_context, trace_id, parent_span_id) = extract_otel_context_from_tcp_headers(headers);
    let x_request_id = headers.get("x-request-id").cloned();
    let x_dynamo_request_id = headers.get("x-dynamo-request-id").cloned();
    let tracestate = headers.get("tracestate").cloned();

    if let (Some(trace_id), Some(parent_id)) = (trace_id.as_ref(), parent_span_id.as_ref()) {
        let span = tracing::info_span!(
            "handle_payload",
            trace_id = trace_id.as_str(),
            parent_id = parent_id.as_str(),
            x_request_id = x_request_id,
            x_dynamo_request_id = x_dynamo_request_id,
            tracestate = tracestate,
            component = component,
            endpoint = endpoint,
            namespace = namespace,
            instance_id = instance_id,
        );

        if let Some(context) = otel_context {
            let _ = span.set_parent(context);
        }
        span
    } else {
        tracing::info_span!(
            "handle_payload",
            x_request_id = x_request_id,
            x_dynamo_request_id = x_dynamo_request_id,
            tracestate = tracestate,
            component = component,
            endpoint = endpoint,
            namespace = namespace,
            instance_id = instance_id,
        )
    }
}

/// Extract OpenTelemetry trace context from TCP/HashMap headers for distributed tracing
fn extract_otel_context_from_tcp_headers(
    headers: &std::collections::HashMap<String, String>,
) -> (
    Option<opentelemetry::Context>,
    Option<String>,
    Option<String>,
) {
    let traceparent_value = match headers.get("traceparent") {
        Some(value) => value.as_str(),
        None => return (None, None, None),
    };

    let (trace_id, parent_span_id) = parse_traceparent(traceparent_value);

    struct TcpHeaderExtractor<'a>(&'a std::collections::HashMap<String, String>);

    impl<'a> Extractor for TcpHeaderExtractor<'a> {
        fn get(&self, key: &str) -> Option<&str> {
            self.0.get(key).map(|s| s.as_str())
        }

        fn keys(&self) -> Vec<&str> {
            vec!["traceparent", "tracestate"]
                .into_iter()
                .filter(|&key| self.0.get(key).is_some())
                .collect()
        }
    }

    let extractor = TcpHeaderExtractor(headers);
    let otel_context = TRACE_PROPAGATOR.extract(&extractor);

    let context_with_trace = if otel_context.span().span_context().is_valid() {
        Some(otel_context)
    } else {
        None
    };

    (context_with_trace, trace_id, parent_span_id)
}

467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
/// Extract OpenTelemetry trace context from NATS headers for distributed tracing
pub fn extract_otel_context_from_nats_headers(
    headers: &async_nats::HeaderMap,
) -> (
    Option<opentelemetry::Context>,
    Option<String>,
    Option<String>,
) {
    let traceparent_value = match headers.get("traceparent") {
        Some(value) => value.as_str(),
        None => return (None, None, None),
    };

    let (trace_id, parent_span_id) = parse_traceparent(traceparent_value);

    struct NatsHeaderExtractor<'a>(&'a async_nats::HeaderMap);

    impl<'a> Extractor for NatsHeaderExtractor<'a> {
        fn get(&self, key: &str) -> Option<&str> {
            self.0.get(key).map(|value| value.as_str())
        }

        fn keys(&self) -> Vec<&str> {
            vec!["traceparent", "tracestate"]
                .into_iter()
                .filter(|&key| self.0.get(key).is_some())
                .collect()
        }
    }

    let extractor = NatsHeaderExtractor(headers);
498
    let otel_context = TRACE_PROPAGATOR.extract(&extractor);
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

    let context_with_trace = if otel_context.span().span_context().is_valid() {
        Some(otel_context)
    } else {
        None
    };

    (context_with_trace, trace_id, parent_span_id)
}

/// Inject OpenTelemetry trace context into NATS headers using W3C Trace Context propagation
pub fn inject_otel_context_into_nats_headers(
    headers: &mut async_nats::HeaderMap,
    context: Option<opentelemetry::Context>,
) {
    let otel_context = context.unwrap_or_else(|| Span::current().context());

    struct NatsHeaderInjector<'a>(&'a mut async_nats::HeaderMap);

    impl<'a> Injector for NatsHeaderInjector<'a> {
        fn set(&mut self, key: &str, value: String) {
            self.0.insert(key, value);
        }
    }

    let mut injector = NatsHeaderInjector(headers);
525
    TRACE_PROPAGATOR.inject_context(&otel_context, &mut injector);
526
527
528
529
530
531
532
}

/// Inject trace context from current span into NATS headers
pub fn inject_current_trace_into_nats_headers(headers: &mut async_nats::HeaderMap) {
    inject_otel_context_into_nats_headers(headers, None);
}

533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
// Inject trace headers into a generic HashMap for HTTP/TCP transports
pub fn inject_trace_headers_into_map(headers: &mut std::collections::HashMap<String, String>) {
    if let Some(trace_context) = get_distributed_tracing_context() {
        // Inject W3C traceparent header
        headers.insert(
            "traceparent".to_string(),
            trace_context.create_traceparent(),
        );

        // Inject optional tracestate
        if let Some(tracestate) = trace_context.tracestate {
            headers.insert("tracestate".to_string(), tracestate);
        }

        // Inject custom request IDs
        if let Some(x_request_id) = trace_context.x_request_id {
            headers.insert("x-request-id".to_string(), x_request_id);
        }
        if let Some(x_dynamo_request_id) = trace_context.x_dynamo_request_id {
            headers.insert("x-dynamo-request-id".to_string(), x_dynamo_request_id);
        }
    }
}

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
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
/// Create a client_request span linked to the parent trace context
pub fn make_client_request_span(
    operation: &str,
    request_id: &str,
    trace_context: Option<&DistributedTraceContext>,
    instance_id: Option<&str>,
) -> Span {
    if let Some(ctx) = trace_context {
        let mut headers = async_nats::HeaderMap::new();
        headers.insert("traceparent", ctx.create_traceparent());

        if let Some(ref tracestate) = ctx.tracestate {
            headers.insert("tracestate", tracestate.as_str());
        }

        let (otel_context, _extracted_trace_id, _extracted_parent_span_id) =
            extract_otel_context_from_nats_headers(&headers);

        let span = if let Some(inst_id) = instance_id {
            tracing::info_span!(
                "client_request",
                operation = operation,
                request_id = request_id,
                instance_id = inst_id,
                trace_id = ctx.trace_id.as_str(),
                parent_id = ctx.span_id.as_str(),
                x_request_id = ctx.x_request_id.as_deref(),
                x_dynamo_request_id = ctx.x_dynamo_request_id.as_deref(),
                // tracestate = ctx.tracestate.as_deref(),
            )
        } else {
            tracing::info_span!(
                "client_request",
                operation = operation,
                request_id = request_id,
                trace_id = ctx.trace_id.as_str(),
                parent_id = ctx.span_id.as_str(),
                x_request_id = ctx.x_request_id.as_deref(),
                x_dynamo_request_id = ctx.x_dynamo_request_id.as_deref(),
                // tracestate = ctx.tracestate.as_deref(),
            )
        };

        if let Some(context) = otel_context {
            let _ = span.set_parent(context);
        }

        span
    } else if let Some(inst_id) = instance_id {
        tracing::info_span!(
            "client_request",
            operation = operation,
            request_id = request_id,
            instance_id = inst_id,
        )
    } else {
        tracing::info_span!(
            "client_request",
            operation = operation,
            request_id = request_id,
        )
    }
619
620
}

621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
#[derive(Debug, Default)]
pub struct FieldVisitor {
    pub fields: HashMap<String, String>,
}

impl Visit for FieldVisitor {
    fn record_str(&mut self, field: &Field, value: &str) {
        self.fields
            .insert(field.name().to_string(), value.to_string());
    }

    fn record_debug(&mut self, field: &Field, value: &dyn std::fmt::Debug) {
        self.fields
            .insert(field.name().to_string(), format!("{:?}", value).to_string());
    }
}

impl<S> Layer<S> for DistributedTraceIdLayer
where
    S: Subscriber + for<'a> tracing_subscriber::registry::LookupSpan<'a>,
{
    // Capture close span time
    // Currently not used but added for future use in timing
    fn on_close(&self, id: Id, ctx: Context<'_, S>) {
        if let Some(span) = ctx.span(&id) {
            let mut extensions = span.extensions_mut();
            if let Some(distributed_tracing_context) =
                extensions.get_mut::<DistributedTraceContext>()
            {
                distributed_tracing_context.end = Some(Instant::now());
            }
        }
    }

655
656
    // Collects span attributes and metadata in on_new_span
    // Final initialization deferred to on_enter when OtelData is available
657
658
659
660
661
662
    fn on_new_span(&self, attrs: &span::Attributes<'_>, id: &Id, ctx: Context<'_, S>) {
        if let Some(span) = ctx.span(id) {
            let mut trace_id: Option<String> = None;
            let mut parent_id: Option<String> = None;
            let mut span_id: Option<String> = None;
            let mut x_request_id: Option<String> = None;
663
            let mut x_dynamo_request_id: Option<String> = None;
664
665
666
667
            let mut tracestate: Option<String> = None;
            let mut visitor = FieldVisitor::default();
            attrs.record(&mut visitor);

668
            // Extract trace_id from span attributes
669
670
671
672
673
674
675
676
            if let Some(trace_id_input) = visitor.fields.get("trace_id") {
                if !is_valid_trace_id(trace_id_input) {
                    tracing::trace!("trace id  '{}' is not valid! Ignoring.", trace_id_input);
                } else {
                    trace_id = Some(trace_id_input.to_string());
                }
            }

677
            // Extract span_id from span attributes
678
679
680
681
682
683
684
685
            if let Some(span_id_input) = visitor.fields.get("span_id") {
                if !is_valid_span_id(span_id_input) {
                    tracing::trace!("span id  '{}' is not valid! Ignoring.", span_id_input);
                } else {
                    span_id = Some(span_id_input.to_string());
                }
            }

686
            // Extract parent_id from span attributes
687
688
689
690
691
692
693
694
            if let Some(parent_id_input) = visitor.fields.get("parent_id") {
                if !is_valid_span_id(parent_id_input) {
                    tracing::trace!("parent id  '{}' is not valid! Ignoring.", parent_id_input);
                } else {
                    parent_id = Some(parent_id_input.to_string());
                }
            }

695
            // Extract tracestate
696
697
698
699
            if let Some(tracestate_input) = visitor.fields.get("tracestate") {
                tracestate = Some(tracestate_input.to_string());
            }

700
            // Extract x_request_id
701
702
703
704
            if let Some(x_request_id_input) = visitor.fields.get("x_request_id") {
                x_request_id = Some(x_request_id_input.to_string());
            }

705
            // Extract x_dynamo_request_id
706
707
708
709
            if let Some(x_request_id_input) = visitor.fields.get("x_dynamo_request_id") {
                x_dynamo_request_id = Some(x_request_id_input.to_string());
            }

710
            // Inherit trace context from parent span if available
711
712
713
714
715
716
717
718
719
            if parent_id.is_none()
                && let Some(parent_span_id) = ctx.current_span().id()
                && let Some(parent_span) = ctx.span(parent_span_id)
            {
                let parent_ext = parent_span.extensions();
                if let Some(parent_tracing_context) = parent_ext.get::<DistributedTraceContext>() {
                    trace_id = Some(parent_tracing_context.trace_id.clone());
                    parent_id = Some(parent_tracing_context.span_id.clone());
                    tracestate = parent_tracing_context.tracestate.clone();
720
721
722
                }
            }

723
            // Validate consistency
724
725
726
727
728
729
730
            if (parent_id.is_some() || span_id.is_some()) && trace_id.is_none() {
                tracing::error!("parent id or span id are set but trace id is not set!");
                // Clear inconsistent IDs to maintain trace integrity
                parent_id = None;
                span_id = None;
            }

731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
            // Store pending context - will be finalized in on_enter
            let mut extensions = span.extensions_mut();
            extensions.insert(PendingDistributedTraceContext {
                trace_id,
                span_id,
                parent_id,
                tracestate,
                x_request_id,
                x_dynamo_request_id,
            });
        }
    }

    // Finalizes the DistributedTraceContext when span is entered
    // At this point, OtelData should have valid trace_id and span_id
    fn on_enter(&self, id: &Id, ctx: Context<'_, S>) {
        if let Some(span) = ctx.span(id) {
            // Check if already initialized (e.g., span re-entered)
            {
                let extensions = span.extensions();
                if extensions.get::<DistributedTraceContext>().is_some() {
                    return;
                }
            }

            // Get the pending context and extract OtelData IDs
            let mut extensions = span.extensions_mut();
            let pending = match extensions.remove::<PendingDistributedTraceContext>() {
                Some(p) => p,
                None => {
                    // This shouldn't happen - on_new_span should have created it
                    tracing::error!("PendingDistributedTraceContext not found in on_enter");
                    return;
                }
            };

            let mut trace_id = pending.trace_id;
            let mut span_id = pending.span_id;
            let parent_id = pending.parent_id;
            let tracestate = pending.tracestate;
            let x_request_id = pending.x_request_id;
            let x_dynamo_request_id = pending.x_dynamo_request_id;

            // Try to extract from OtelData if not already set
            // Need to drop extensions_mut to get immutable borrow for OtelData
            drop(extensions);

            if trace_id.is_none() || span_id.is_none() {
                let extensions = span.extensions();
                if let Some(otel_data) = extensions.get::<tracing_opentelemetry::OtelData>() {
                    // Extract trace_id from OTEL data if not already set
                    if trace_id.is_none()
                        && let Some(otel_trace_id) = otel_data.trace_id()
                    {
                        let trace_id_str = format!("{}", otel_trace_id);
                        if is_valid_trace_id(&trace_id_str) {
                            trace_id = Some(trace_id_str);
                        }
                    }

                    // Extract span_id from OTEL data if not already set
                    if span_id.is_none()
                        && let Some(otel_span_id) = otel_data.span_id()
                    {
                        let span_id_str = format!("{}", otel_span_id);
                        if is_valid_span_id(&span_id_str) {
                            span_id = Some(span_id_str);
                        }
                    }
                }
            }

            // Panic if we still don't have required IDs
804
            if trace_id.is_none() {
805
806
807
                panic!(
                    "trace_id is not set in on_enter - OtelData may not be properly initialized"
                );
808
            }
809

810
            if span_id.is_none() {
811
                panic!("span_id is not set in on_enter - OtelData may not be properly initialized");
812
813
            }

814
            // Re-acquire mutable borrow to insert the finalized context
815
816
817
818
819
820
            let mut extensions = span.extensions_mut();
            extensions.insert(DistributedTraceContext {
                trace_id: trace_id.expect("Trace ID must be set"),
                span_id: span_id.expect("Span ID must be set"),
                parent_id,
                tracestate,
821
                start: Some(Instant::now()),
822
823
                end: None,
                x_request_id,
824
                x_dynamo_request_id,
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
            });
        }
    }
}

// Enables functions to retreive their current
// context for adding to distributed headers
pub fn get_distributed_tracing_context() -> Option<DistributedTraceContext> {
    Span::current()
        .with_subscriber(|(id, subscriber)| {
            subscriber
                .downcast_ref::<Registry>()
                .and_then(|registry| registry.span_data(id))
                .and_then(|span_data| {
                    let extensions = span_data.extensions();
                    extensions.get::<DistributedTraceContext>().cloned()
                })
        })
        .flatten()
}

846
/// Initialize the logger - must be called when Tokio runtime is available
847
pub fn init() {
848
849
850
851
852
853
    INIT.call_once(|| {
        if let Err(e) = setup_logging() {
            eprintln!("Failed to initialize logging: {}", e);
            std::process::exit(1);
        }
    });
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
}

#[cfg(feature = "tokio-console")]
fn setup_logging() {
    let tokio_console_layer = console_subscriber::ConsoleLayer::builder()
        .with_default_env()
        .server_addr(([0, 0, 0, 0], console_subscriber::Server::DEFAULT_PORT))
        .spawn();
    let tokio_console_target = tracing_subscriber::filter::Targets::new()
        .with_default(LevelFilter::ERROR)
        .with_target("runtime", LevelFilter::TRACE)
        .with_target("tokio", LevelFilter::TRACE);
    let l = fmt::layer()
        .with_ansi(!disable_ansi_logging())
        .event_format(fmt::format().compact().with_timer(TimeFormatter::new()))
        .with_writer(std::io::stderr)
870
        .with_filter(filters(load_config()));
871
872
873
874
875
876
877
    tracing_subscriber::registry()
        .with(l)
        .with(tokio_console_layer.with_filter(tokio_console_target))
        .init();
}

#[cfg(not(feature = "tokio-console"))]
878
fn setup_logging() -> Result<(), Box<dyn std::error::Error>> {
879
880
    let fmt_filter_layer = filters(load_config());
    let trace_filter_layer = filters(load_config());
881
882
    let otel_filter_layer = filters(load_config());

883
884
885
886
887
    if jsonl_logging_enabled() {
        let l = fmt::layer()
            .with_ansi(false)
            .event_format(CustomJsonFormatter::new())
            .with_writer(std::io::stderr)
888
            .with_filter(fmt_filter_layer);
889
890
891
892
893

        // Create OpenTelemetry tracer - conditionally export to OTLP based on env var
        let service_name = get_service_name();

        // Build tracer provider - with or without OTLP export
894
        let (tracer_provider, endpoint_opt) = if otlp_exporter_enabled() {
895
            // Export enabled: create OTLP exporter with batch processor
896
            let endpoint = std::env::var(env_logging::otlp::OTEL_EXPORTER_OTLP_TRACES_ENDPOINT)
897
898
899
900
901
                .unwrap_or_else(|_| DEFAULT_OTLP_ENDPOINT.to_string());

            // Initialize OTLP exporter using gRPC (Tonic)
            let otlp_exporter = opentelemetry_otlp::SpanExporter::builder()
                .with_tonic()
902
                .with_endpoint(&endpoint)
903
904
905
                .build()?;

            // Create tracer provider with batch exporter and service name
906
            let provider = opentelemetry_sdk::trace::SdkTracerProvider::builder()
907
908
909
910
911
912
                .with_batch_exporter(otlp_exporter)
                .with_resource(
                    opentelemetry_sdk::Resource::builder_empty()
                        .with_service_name(service_name.clone())
                        .build(),
                )
913
914
915
                .build();

            (provider, Some(endpoint))
916
917
        } else {
            // No export - traces generated locally only (for logging/trace IDs)
918
            let provider = opentelemetry_sdk::trace::SdkTracerProvider::builder()
919
920
921
922
923
                .with_resource(
                    opentelemetry_sdk::Resource::builder_empty()
                        .with_service_name(service_name.clone())
                        .build(),
                )
924
925
926
                .build();

            (provider, None)
927
928
929
        };

        // Get a tracer from the provider
930
        let tracer = tracer_provider.tracer(service_name.clone());
931

932
        tracing_subscriber::registry()
933
934
935
936
937
            .with(
                tracing_opentelemetry::layer()
                    .with_tracer(tracer)
                    .with_filter(otel_filter_layer),
            )
938
            .with(DistributedTraceIdLayer.with_filter(trace_filter_layer))
939
940
            .with(l)
            .init();
941
942
943
944
945
946
947
948
949
950
951
952
953
954

        // Log initialization status after subscriber is ready
        if let Some(endpoint) = endpoint_opt {
            tracing::info!(
                endpoint = %endpoint,
                service = %service_name,
                "OpenTelemetry OTLP export enabled"
            );
        } else {
            tracing::info!(
                service = %service_name,
                "OpenTelemetry OTLP export disabled, traces local only"
            );
        }
955
956
957
958
959
    } else {
        let l = fmt::layer()
            .with_ansi(!disable_ansi_logging())
            .event_format(fmt::format().compact().with_timer(TimeFormatter::new()))
            .with_writer(std::io::stderr)
960
            .with_filter(fmt_filter_layer);
961

962
963
        tracing_subscriber::registry().with(l).init();
    }
964
965

    Ok(())
966
967
968
969
970
}

fn filters(config: LoggingConfig) -> EnvFilter {
    let mut filter_layer = EnvFilter::builder()
        .with_default_directive(config.log_level.parse().unwrap())
971
        .with_env_var(env_logging::DYN_LOG)
972
973
974
975
976
977
978
979
980
        .from_env_lossy();

    for (module, level) in config.log_filters {
        match format!("{module}={level}").parse::<Directive>() {
            Ok(d) => {
                filter_layer = filter_layer.add_directive(d);
            }
            Err(e) => {
                eprintln!("Failed parsing filter '{level}' for module '{module}': {e}");
981
982
            }
        }
983
984
    }
    filter_layer
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
}

/// Log a message with file and line info
/// Used by Python wrapper
pub fn log_message(level: &str, message: &str, module: &str, file: &str, line: u32) {
    let level = match level {
        "debug" => log::Level::Debug,
        "info" => log::Level::Info,
        "warn" => log::Level::Warn,
        "error" => log::Level::Error,
        "warning" => log::Level::Warn,
        _ => log::Level::Info,
    };
    log::logger().log(
        &log::Record::builder()
            .args(format_args!("{}", message))
            .level(level)
            .target(module)
            .file(Some(file))
            .line(Some(line))
            .build(),
    );
}

fn load_config() -> LoggingConfig {
1010
1011
    let config_path =
        std::env::var(env_logging::DYN_LOGGING_CONFIG_PATH).unwrap_or_else(|_| "".to_string());
1012
1013
    let figment = Figment::new()
        .merge(Serialized::defaults(LoggingConfig::default()))
Neelay Shah's avatar
Neelay Shah committed
1014
        .merge(Toml::file("/opt/dynamo/etc/logging.toml"))
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
        .merge(Toml::file(config_path));

    figment.extract().unwrap()
}

#[derive(Serialize)]
struct JsonLog<'a> {
    time: String,
    level: String,
    #[serde(skip_serializing_if = "Option::is_none")]
1025
    file: Option<&'a str>,
1026
    #[serde(skip_serializing_if = "Option::is_none")]
1027
1028
    line: Option<u32>,
    target: &'a str,
1029
1030
1031
1032
1033
    message: serde_json::Value,
    #[serde(flatten)]
    fields: BTreeMap<String, serde_json::Value>,
}

Ryan Olson's avatar
Ryan Olson committed
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
struct TimeFormatter {
    use_local_tz: bool,
}

impl TimeFormatter {
    fn new() -> Self {
        Self {
            use_local_tz: crate::config::use_local_timezone(),
        }
    }

    fn format_now(&self) -> String {
        if self.use_local_tz {
            chrono::Local::now()
1048
                .format("%Y-%m-%dT%H:%M:%S%.6f%:z")
Ryan Olson's avatar
Ryan Olson committed
1049
1050
1051
                .to_string()
        } else {
            chrono::Utc::now()
1052
                .format("%Y-%m-%dT%H:%M:%S%.6fZ")
Ryan Olson's avatar
Ryan Olson committed
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
                .to_string()
        }
    }
}

impl FormatTime for TimeFormatter {
    fn format_time(&self, w: &mut fmt::format::Writer<'_>) -> std::fmt::Result {
        write!(w, "{}", self.format_now())
    }
}

struct CustomJsonFormatter {
    time_formatter: TimeFormatter,
}

impl CustomJsonFormatter {
    fn new() -> Self {
        Self {
            time_formatter: TimeFormatter::new(),
        }
    }
}
1075

1076
1077
use once_cell::sync::Lazy;
use regex::Regex;
1078
1079
1080
1081
1082

/// Static W3C Trace Context propagator instance to avoid repeated allocations
static TRACE_PROPAGATOR: Lazy<opentelemetry_sdk::propagation::TraceContextPropagator> =
    Lazy::new(opentelemetry_sdk::propagation::TraceContextPropagator::new);

1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
fn parse_tracing_duration(s: &str) -> Option<u64> {
    static RE: Lazy<Regex> =
        Lazy::new(|| Regex::new(r#"^["']?\s*([0-9.]+)\s*(µs|us|ns|ms|s)\s*["']?$"#).unwrap());
    let captures = RE.captures(s)?;
    let value: f64 = captures[1].parse().ok()?;
    let unit = &captures[2];
    match unit {
        "ns" => Some((value / 1000.0) as u64),
        "µs" | "us" => Some(value as u64),
        "ms" => Some((value * 1000.0) as u64),
        "s" => Some((value * 1_000_000.0) as u64),
        _ => None,
    }
}

1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
impl<S, N> tracing_subscriber::fmt::FormatEvent<S, N> for CustomJsonFormatter
where
    S: Subscriber + for<'a> LookupSpan<'a>,
    N: for<'a> FormatFields<'a> + 'static,
{
    fn format_event(
        &self,
        ctx: &FmtContext<'_, S, N>,
        mut writer: Writer<'_>,
        event: &Event<'_>,
    ) -> std::fmt::Result {
        let mut visitor = JsonVisitor::default();
1110
        let time = self.time_formatter.format_now();
1111
        event.record(&mut visitor);
1112
        let mut message = visitor
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
            .fields
            .remove("message")
            .unwrap_or(serde_json::Value::String("".to_string()));

        let current_span = event
            .parent()
            .and_then(|id| ctx.span(id))
            .or_else(|| ctx.lookup_current());
        if let Some(span) = current_span {
            let ext = span.extensions();
            let data = ext.get::<FormattedFields<N>>().unwrap();
            let span_fields: Vec<(&str, &str)> = data
                .fields
                .split(' ')
                .filter_map(|entry| entry.split_once('='))
                .collect();
            for (name, value) in span_fields {
                visitor.fields.insert(
                    name.to_string(),
                    serde_json::Value::String(value.trim_matches('"').to_string()),
                );
            }
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165

            let busy_us = visitor
                .fields
                .remove("time.busy")
                .and_then(|v| parse_tracing_duration(&v.to_string()));
            let idle_us = visitor
                .fields
                .remove("time.idle")
                .and_then(|v| parse_tracing_duration(&v.to_string()));

            if let (Some(busy_us), Some(idle_us)) = (busy_us, idle_us) {
                visitor.fields.insert(
                    "time.busy_us".to_string(),
                    serde_json::Value::Number(busy_us.into()),
                );
                visitor.fields.insert(
                    "time.idle_us".to_string(),
                    serde_json::Value::Number(idle_us.into()),
                );
                visitor.fields.insert(
                    "time.duration_us".to_string(),
                    serde_json::Value::Number((busy_us + idle_us).into()),
                );
            }

            message = match message.as_str() {
                Some("new") => serde_json::Value::String("SPAN_CREATED".to_string()),
                Some("close") => serde_json::Value::String("SPAN_CLOSED".to_string()),
                _ => message.clone(),
            };

1166
1167
1168
1169
1170
            visitor.fields.insert(
                "span_name".to_string(),
                serde_json::Value::String(span.name().to_string()),
            );

1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
            if let Some(tracing_context) = ext.get::<DistributedTraceContext>() {
                visitor.fields.insert(
                    "span_id".to_string(),
                    serde_json::Value::String(tracing_context.span_id.clone()),
                );
                visitor.fields.insert(
                    "trace_id".to_string(),
                    serde_json::Value::String(tracing_context.trace_id.clone()),
                );
                if let Some(parent_id) = tracing_context.parent_id.clone() {
                    visitor.fields.insert(
                        "parent_id".to_string(),
                        serde_json::Value::String(parent_id),
                    );
                } else {
                    visitor.fields.remove("parent_id");
                }
                if let Some(tracestate) = tracing_context.tracestate.clone() {
                    visitor.fields.insert(
                        "tracestate".to_string(),
                        serde_json::Value::String(tracestate),
                    );
                } else {
                    visitor.fields.remove("tracestate");
                }
                if let Some(x_request_id) = tracing_context.x_request_id.clone() {
                    visitor.fields.insert(
                        "x_request_id".to_string(),
                        serde_json::Value::String(x_request_id),
                    );
                } else {
                    visitor.fields.remove("x_request_id");
                }
1204
1205
1206
1207
1208
1209
1210
1211
1212

                if let Some(x_dynamo_request_id) = tracing_context.x_dynamo_request_id.clone() {
                    visitor.fields.insert(
                        "x_dynamo_request_id".to_string(),
                        serde_json::Value::String(x_dynamo_request_id),
                    );
                } else {
                    visitor.fields.remove("x_dynamo_request_id");
                }
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
            } else {
                tracing::error!(
                    "Distributed Trace Context not found, falling back to internal ids"
                );
                visitor.fields.insert(
                    "span_id".to_string(),
                    serde_json::Value::String(span.id().into_u64().to_string()),
                );
                if let Some(parent) = span.parent() {
                    visitor.fields.insert(
                        "parent_id".to_string(),
                        serde_json::Value::String(parent.id().into_u64().to_string()),
                    );
                }
            }
        } else {
            let reserved_fields = [
                "trace_id",
                "span_id",
                "parent_id",
                "span_name",
                "tracestate",
            ];
            for reserved_field in reserved_fields {
                visitor.fields.remove(reserved_field);
            }
        }
1240
1241
1242
        let metadata = event.metadata();
        let log = JsonLog {
            level: metadata.level().to_string(),
1243
1244
1245
1246
            time,
            file: metadata.file(),
            line: metadata.line(),
            target: metadata.target(),
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
            message,
            fields: visitor.fields,
        };
        let json = serde_json::to_string(&log).unwrap();
        writeln!(writer, "{json}")
    }
}

#[derive(Default)]
struct JsonVisitor {
    fields: BTreeMap<String, serde_json::Value>,
}

impl tracing::field::Visit for JsonVisitor {
    fn record_debug(&mut self, field: &tracing::field::Field, value: &dyn std::fmt::Debug) {
        self.fields.insert(
            field.name().to_string(),
            serde_json::Value::String(format!("{value:?}")),
        );
    }

    fn record_str(&mut self, field: &tracing::field::Field, value: &str) {
1269
1270
1271
1272
1273
1274
1275
1276
        if field.name() != "message" {
            match serde_json::from_str::<Value>(value) {
                Ok(json_val) => self.fields.insert(field.name().to_string(), json_val),
                Err(_) => self.fields.insert(field.name().to_string(), value.into()),
            };
        } else {
            self.fields.insert(field.name().to_string(), value.into());
        }
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
    }

    fn record_bool(&mut self, field: &tracing::field::Field, value: bool) {
        self.fields
            .insert(field.name().to_string(), serde_json::Value::Bool(value));
    }

    fn record_i64(&mut self, field: &tracing::field::Field, value: i64) {
        self.fields.insert(
            field.name().to_string(),
            serde_json::Value::Number(value.into()),
        );
    }

    fn record_u64(&mut self, field: &tracing::field::Field, value: u64) {
        self.fields.insert(
            field.name().to_string(),
            serde_json::Value::Number(value.into()),
        );
    }

    fn record_f64(&mut self, field: &tracing::field::Field, value: f64) {
        use serde_json::value::Number;
        self.fields.insert(
            field.name().to_string(),
            serde_json::Value::Number(Number::from_f64(value).unwrap_or(0.into())),
        );
    }
}
1306
1307
1308
1309

#[cfg(test)]
pub mod tests {
    use super::*;
1310
    use anyhow::{Result, anyhow};
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
    use chrono::{DateTime, Utc};
    use jsonschema::{Draft, JSONSchema};
    use serde_json::Value;
    use std::fs::File;
    use std::io::{BufRead, BufReader};
    use stdio_override::*;
    use tempfile::NamedTempFile;

    static LOG_LINE_SCHEMA: &str = r#"
    {
      "$schema": "http://json-schema.org/draft-07/schema#",
      "title": "Runtime Log Line",
      "type": "object",
      "required": [
        "file",
        "level",
        "line",
        "message",
        "target",
        "time"
      ],
      "properties": {
        "file":      { "type": "string" },
        "level":     { "type": "string", "enum": ["ERROR", "WARN", "INFO", "DEBUG", "TRACE"] },
        "line":      { "type": "integer" },
        "message":   { "type": "string" },
        "target":    { "type": "string" },
        "time":      { "type": "string", "format": "date-time" },
        "span_id":   { "type": "string", "pattern": "^[a-f0-9]{16}$" },
        "parent_id": { "type": "string", "pattern": "^[a-f0-9]{16}$" },
        "trace_id":  { "type": "string", "pattern": "^[a-f0-9]{32}$" },
        "span_name": { "type": "string" },
        "time.busy_us":     { "type": "integer" },
        "time.duration_us": { "type": "integer" },
        "time.idle_us":     { "type": "integer" },
        "tracestate": { "type": "string" }
      },
      "additionalProperties": true
    }
    "#;

1352
    #[tracing::instrument(skip_all)]
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
    async fn parent() {
        tracing::trace!(message = "parent!");
        if let Some(my_ctx) = get_distributed_tracing_context() {
            tracing::info!(my_trace_id = my_ctx.trace_id);
        }
        child().await;
    }

    #[tracing::instrument(skip_all)]
    async fn child() {
        tracing::trace!(message = "child");
        if let Some(my_ctx) = get_distributed_tracing_context() {
            tracing::info!(my_trace_id = my_ctx.trace_id);
        }
        grandchild().await;
    }

    #[tracing::instrument(skip_all)]
    async fn grandchild() {
        tracing::trace!(message = "grandchild");
        if let Some(my_ctx) = get_distributed_tracing_context() {
            tracing::info!(my_trace_id = my_ctx.trace_id);
        }
    }

    pub fn load_log(file_name: &str) -> Result<Vec<serde_json::Value>> {
        let schema_json: Value =
            serde_json::from_str(LOG_LINE_SCHEMA).expect("schema parse failure");
        let compiled_schema = JSONSchema::options()
            .with_draft(Draft::Draft7)
            .compile(&schema_json)
            .expect("Invalid schema");

        let f = File::open(file_name)?;
        let reader = BufReader::new(f);
        let mut result = Vec::new();

        for (line_num, line) in reader.lines().enumerate() {
            let line = line?;
            let val: Value = serde_json::from_str(&line)
                .map_err(|e| anyhow!("Line {}: invalid JSON: {}", line_num + 1, e))?;

            if let Err(errors) = compiled_schema.validate(&val) {
                let errs = errors.map(|e| e.to_string()).collect::<Vec<_>>().join("; ");
                return Err(anyhow!(
                    "Line {}: JSON Schema Validation errors: {}",
                    line_num + 1,
                    errs
                ));
            }
            println!("{}", val);
            result.push(val);
        }
        Ok(result)
    }

    #[tokio::test]
    async fn test_json_log_capture() -> Result<()> {
        #[allow(clippy::redundant_closure_call)]
        let _ = temp_env::async_with_vars(
1413
            [(env_logging::DYN_LOGGING_JSONL, Some("1"))],
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
            (async || {
                let tmp_file = NamedTempFile::new().unwrap();
                let file_name = tmp_file.path().to_str().unwrap();
                let guard = StderrOverride::from_file(file_name)?;
                init();
                parent().await;
                drop(guard);

                let lines = load_log(file_name)?;

1424
1425
                // 1. Extract the dynamically generated trace ID and validate consistency
                // All logs should have the same trace_id since they're part of the same trace
1426
                // Skip any initialization logs that don't have trace_id (e.g., OTLP setup messages)
1427
1428
1429
1430
1431
                //
                // Note: This test can fail if logging was already initialized by another test running
                // in parallel. Logging initialization is global (Once) and can only happen once per process.
                // If no trace_id is found, skip validation gracefully.
                let Some(trace_id) = lines
1432
1433
                    .iter()
                    .find_map(|log_line| log_line.get("trace_id").and_then(|v| v.as_str()))
1434
1435
1436
1437
1438
                    .map(|s| s.to_string())
                else {
                    // Skip test if logging was already initialized - we can't control the output format
                    return Ok(());
                };
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461

                // Verify trace_id is not a zero/invalid ID
                assert_ne!(
                    trace_id, "00000000000000000000000000000000",
                    "trace_id should not be a zero/invalid ID"
                );
                assert!(
                    !trace_id.chars().all(|c| c == '0'),
                    "trace_id should not be all zeros"
                );

                // Verify all logs have the same trace_id
                for log_line in &lines {
                    if let Some(line_trace_id) = log_line.get("trace_id") {
                        assert_eq!(
                            line_trace_id.as_str().unwrap(),
                            &trace_id,
                            "All logs should have the same trace_id"
                        );
                    }
                }

                // Validate my_trace_id matches the actual trace ID
1462
1463
1464
1465
                for log_line in &lines {
                    if let Some(my_trace_id) = log_line.get("my_trace_id") {
                        assert_eq!(
                            my_trace_id,
1466
1467
                            &serde_json::Value::String(trace_id.clone()),
                            "my_trace_id should match the trace_id from distributed tracing context"
1468
1469
1470
1471
                        );
                    }
                }

1472
1473
1474
                // 2. Validate span IDs exist and are properly formatted
                let mut span_ids_seen: std::collections::HashSet<String> = std::collections::HashSet::new();
                let mut span_timestamps: std::collections::HashMap<String, DateTime<Utc>> = std::collections::HashMap::new();
1475
1476

                for log_line in &lines {
1477
1478
1479
1480
1481
1482
1483
1484
                    if let Some(span_id) = log_line.get("span_id") {
                        let span_id_str = span_id.as_str().unwrap();
                        assert!(
                            is_valid_span_id(span_id_str),
                            "Invalid span_id format: {}",
                            span_id_str
                        );
                        span_ids_seen.insert(span_id_str.to_string());
1485
1486
                    }

1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
                    // Validate timestamp format and track span timestamps
                    if let Some(time_str) = log_line.get("time").and_then(|v| v.as_str()) {
                        let timestamp = DateTime::parse_from_rfc3339(time_str)
                            .expect("All timestamps should be valid RFC3339 format")
                            .with_timezone(&Utc);

                        // Track timestamp for each span_name
                        if let Some(span_name) = log_line.get("span_name").and_then(|v| v.as_str()) {
                            span_timestamps.insert(span_name.to_string(), timestamp);
                        }
                    }
1498
1499
                }

1500
1501
                // 3. Validate parent-child span relationships
                // Extract span IDs for each span by looking at their log messages
1502
1503
1504
                let parent_span_id = lines
                    .iter()
                    .find(|log_line| {
1505
1506
                        log_line.get("span_name")
                            .and_then(|v| v.as_str()) == Some("parent")
1507
1508
                    })
                    .and_then(|log_line| {
1509
1510
1511
                        log_line.get("span_id")
                            .and_then(|v| v.as_str())
                            .map(|s| s.to_string())
1512
                    })
1513
                    .expect("Should find parent span with span_id");
1514
1515
1516
1517

                let child_span_id = lines
                    .iter()
                    .find(|log_line| {
1518
1519
                        log_line.get("span_name")
                            .and_then(|v| v.as_str()) == Some("child")
1520
1521
                    })
                    .and_then(|log_line| {
1522
1523
1524
                        log_line.get("span_id")
                            .and_then(|v| v.as_str())
                            .map(|s| s.to_string())
1525
                    })
1526
                    .expect("Should find child span with span_id");
1527

1528
                let grandchild_span_id = lines
1529
1530
                    .iter()
                    .find(|log_line| {
1531
1532
                        log_line.get("span_name")
                            .and_then(|v| v.as_str()) == Some("grandchild")
1533
1534
                    })
                    .and_then(|log_line| {
1535
1536
1537
                        log_line.get("span_id")
                            .and_then(|v| v.as_str())
                            .map(|s| s.to_string())
1538
                    })
1539
1540
1541
1542
1543
1544
                    .expect("Should find grandchild span with span_id");

                // Verify span IDs are unique
                assert_ne!(parent_span_id, child_span_id, "Parent and child should have different span IDs");
                assert_ne!(child_span_id, grandchild_span_id, "Child and grandchild should have different span IDs");
                assert_ne!(parent_span_id, grandchild_span_id, "Parent and grandchild should have different span IDs");
1545

1546
                // Verify parent span has no parent_id
1547
                for log_line in &lines {
1548
1549
1550
1551
                    if let Some(span_name) = log_line.get("span_name")
                        && let Some(span_name_str) = span_name.as_str()
                        && span_name_str == "parent"
                    {
1552
1553
1554
1555
                        assert!(
                            log_line.get("parent_id").is_none(),
                            "Parent span should not have a parent_id"
                        );
1556
1557
1558
                    }
                }

1559
                // Verify child span's parent_id is parent_span_id
1560
                for log_line in &lines {
1561
1562
1563
1564
                    if let Some(span_name) = log_line.get("span_name")
                        && let Some(span_name_str) = span_name.as_str()
                        && span_name_str == "child"
                    {
1565
1566
1567
                        let parent_id = log_line.get("parent_id")
                            .and_then(|v| v.as_str())
                            .expect("Child span should have a parent_id");
1568
                        assert_eq!(
1569
1570
1571
                            parent_id,
                            parent_span_id,
                            "Child's parent_id should match parent's span_id"
1572
                        );
1573
1574
1575
                    }
                }

1576
                // Verify grandchild span's parent_id is child_span_id
1577
                for log_line in &lines {
1578
1579
1580
1581
                    if let Some(span_name) = log_line.get("span_name")
                        && let Some(span_name_str) = span_name.as_str()
                        && span_name_str == "grandchild"
                    {
1582
1583
1584
                        let parent_id = log_line.get("parent_id")
                            .and_then(|v| v.as_str())
                            .expect("Grandchild span should have a parent_id");
1585
                        assert_eq!(
1586
1587
1588
                            parent_id,
                            child_span_id,
                            "Grandchild's parent_id should match child's span_id"
1589
                        );
1590
1591
1592
                    }
                }

1593
1594
1595
1596
1597
1598
1599
                // 4. Validate timestamp ordering - spans should log in execution order
                let parent_time = span_timestamps.get("parent")
                    .expect("Should have timestamp for parent span");
                let child_time = span_timestamps.get("child")
                    .expect("Should have timestamp for child span");
                let grandchild_time = span_timestamps.get("grandchild")
                    .expect("Should have timestamp for grandchild span");
1600

1601
                // Parent logs first (or at same time), then child, then grandchild
1602
                assert!(
1603
1604
1605
1606
                    parent_time <= child_time,
                    "Parent span should log before or at same time as child span (parent: {}, child: {})",
                    parent_time,
                    child_time
1607
1608
                );
                assert!(
1609
1610
1611
1612
                    child_time <= grandchild_time,
                    "Child span should log before or at same time as grandchild span (child: {}, grandchild: {})",
                    child_time,
                    grandchild_time
1613
1614
1615
1616
1617
1618
1619
1620
1621
                );

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