feat: migrate from legacy jaeger to native otlp

binaries/daemon/Cargo.toml

@@ -20,8 +20,8 @@ telemetry = ["dep:tracing-opentelemetry"]
 eyre = "0.6.8"
 tokio = { version = "1.20.1", features = ["full"] }
 tokio-stream = { version = "0.1.11", features = ["net"] }
-tracing = "0.1.36"
-tracing-opentelemetry = { version = "0.18.0", optional = true }
+tracing = "0.1.41"
+tracing-opentelemetry = { version = "0.28.0", optional = true }
 futures-concurrency = "7.1.0"
 serde_json = "1.0.86"
 dora-core = { workspace = true }

binaries/runtime/Cargo.toml

@@ -26,10 +26,10 @@ tokio = { version = "1.24.2", features = ["full"] }
 tokio-stream = "0.1.8"
 # pyo3-abi3 flag allow simpler linking. See: https://pyo3.rs/v0.13.2/building_and_distribution.html
 pyo3 = { workspace = true, features = ["eyre", "abi3-py37"], optional = true }
-tracing = "0.1.36"
+tracing = "0.1.41"
 dora-download = { workspace = true }
 flume = "0.10.14"
-tracing-opentelemetry = { version = "0.18.0", optional = true }
+tracing-opentelemetry = { version = "0.28.0", optional = true }
 pythonize = { workspace = true, optional = true }
 arrow = { workspace = true, features = ["ffi"] }
 aligned-vec = "0.5.0"

libraries/extensions/telemetry/metrics/Cargo.toml

@@ -10,8 +10,9 @@ repository.workspace = true
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
 [dependencies]
-opentelemetry = { version = "0.22.0", features = ["metrics"] }
-opentelemetry-otlp = { version = "0.15.0", features = ["tonic", "metrics"] }
-opentelemetry_sdk = { version = "0.22.0", features = ["rt-tokio", "metrics"] }
+opentelemetry = { version = "0.27.1", features = ["metrics"] }
+opentelemetry-otlp = { version = "0.27.0", features = ["tonic", "metrics"] }
+opentelemetry_sdk = { version = "0.27.1", features = ["rt-tokio", "metrics"] }
 eyre = "0.6.12"
-opentelemetry-system-metrics = { version = "0.1.8" }
+# TODO: change to crate new version once https://github.com/haixuanTao/opentelemetry-system-metrics/pull/7 is merged
+opentelemetry-system-metrics = { git = "https://github.com/wiseaidev/opentelemetry-system-metrics", branch = "opentelemetry" }

libraries/extensions/telemetry/metrics/src/lib.rs

@@ -13,37 +13,45 @@
 use std::time::Duration;
 
 use eyre::{Context, Result};
-use opentelemetry::metrics::{self, MeterProvider as _};
-use opentelemetry_otlp::{ExportConfig, WithExportConfig};
-use opentelemetry_sdk::{metrics::SdkMeterProvider, runtime};
+use opentelemetry::metrics::MeterProvider;
+use opentelemetry_otlp::Protocol;
+use opentelemetry_otlp::WithExportConfig;
+use opentelemetry_sdk::metrics::{MetricResult, SdkMeterProvider};
 use opentelemetry_system_metrics::init_process_observer;
 /// Init opentelemetry meter
 ///
 /// Use the default Opentelemetry exporter with default config
 /// TODO: Make Opentelemetry configurable
 ///
-pub fn init_metrics() -> metrics::Result<SdkMeterProvider> {
+pub fn init_metrics() -> MetricResult<SdkMeterProvider> {
     let endpoint = std::env::var("OTEL_EXPORTER_OTLP_METRICS_ENDPOINT")
-        .unwrap_or_else(|_| "http://localhost:4317".to_string());
-    let export_config = ExportConfig {
-        endpoint,
-        ..ExportConfig::default()
-    };
+        .unwrap_or_else(|_| "http://localhost:4318/v1/metrics".to_string());
 
-    opentelemetry_otlp::new_pipeline()
-        .metrics(runtime::Tokio)
-        .with_exporter(
-            opentelemetry_otlp::new_exporter()
-                .tonic()
-                .with_export_config(export_config),
-        )
-        .with_period(Duration::from_secs(10))
+    let exporter = opentelemetry_otlp::MetricExporter::builder()
+        .with_tonic()
+        .with_endpoint(endpoint)
+        .with_protocol(Protocol::Grpc)
+        .with_timeout(Duration::from_secs(10))
         .build()
+        .unwrap();
+
+    let reader = opentelemetry_sdk::metrics::PeriodicReader::builder(
+        exporter,
+        opentelemetry_sdk::runtime::Tokio,
+    )
+    .with_interval(std::time::Duration::from_secs(10))
+    .with_timeout(Duration::from_secs(10))
+    .build();
+
+    Ok(opentelemetry_sdk::metrics::SdkMeterProvider::builder()
+        .with_reader(reader)
+        .build())
 }
 
 pub fn init_meter_provider(meter_id: String) -> Result<SdkMeterProvider> {
+    let meter_id: &'static str = Box::leak(meter_id.into_boxed_str());
     let meter_provider = init_metrics().context("Could not create opentelemetry meter")?;
     let meter = meter_provider.meter(meter_id);
-    init_process_observer(meter).context("could not initiate system metrics observer")?;
+    init_process_observer(meter.clone()).context("could not initiate system metrics observer")?;
     Ok(meter_provider)
 }

libraries/extensions/telemetry/tracing/Cargo.toml

@@ -13,8 +13,10 @@ repository.workspace = true
 
 [dependencies]
 tracing-subscriber = { version = "0.3.15", features = ["env-filter"] }
-tracing-opentelemetry = { version = "0.18.0" }
+tracing-opentelemetry = { version = "0.28.0" }
 eyre = "0.6.8"
-tracing = "0.1.36"
-opentelemetry = { version = "0.18.0", features = ["rt-tokio", "metrics"] }
-opentelemetry-jaeger = { version = "0.17.0", features = ["rt-tokio"] }
+tracing = "0.1.41"
+opentelemetry = { version = "0.27.1" }
+opentelemetry-otlp = { version = "0.27.0", default-features = false, features = ["http-proto", "http-json", "logs", "internal-logs"] }
+opentelemetry_sdk = { version = "0.27.1", features = ["rt-tokio", "logs"]}
+opentelemetry-appender-tracing = { version = "0.27.0"}

libraries/extensions/telemetry/tracing/src/lib.rs

@@ -12,7 +12,7 @@ use tracing_subscriber::{
 };
 
 use eyre::ContextCompat;
-use tracing_subscriber::Registry;
+use opentelemetry_appender_tracing::layer::OpenTelemetryTracingBridge;
 pub mod telemetry;
 
 pub fn set_up_tracing(name: &str) -> eyre::Result<()> {
@@ -62,13 +62,12 @@ pub fn set_up_tracing_opts(
         let endpoint = endpoint
             .to_str()
             .wrap_err("Could not parse env variable: DORA_JAEGER_TRACING")?;
-        let tracer = crate::telemetry::init_jaeger_tracing(name, endpoint)
+        let tracer = crate::telemetry::init_jaeger_tracing(name.to_string(), endpoint)
             .wrap_err("Could not instantiate tracing")?;
-        let telemetry = tracing_opentelemetry::layer().with_tracer(tracer);
-        layers.push(telemetry.boxed());
+        layers.push(Box::new(OpenTelemetryTracingBridge::new(&tracer)));
     }
 
-    let registry = Registry::default().with(layers);
+    let registry = tracing_subscriber::registry().with(layers);
     tracing::subscriber::set_global_default(registry).context(format!(
         "failed to set tracing global subscriber for {name}"
     ))

libraries/extensions/telemetry/tracing/src/telemetry.rs

@@ -1,7 +1,9 @@
 use opentelemetry::propagation::Extractor;
-use opentelemetry::sdk::{propagation::TraceContextPropagator, trace as sdktrace};
-use opentelemetry::trace::TraceError;
 use opentelemetry::{global, Context};
+use opentelemetry::{Key, KeyValue, Value};
+use opentelemetry_otlp::WithExportConfig;
+use opentelemetry_otlp::{LogExporter, Protocol};
+use opentelemetry_sdk::{logs::LoggerProvider, Resource};
 use std::collections::HashMap;
 
 struct MetadataMap<'a>(HashMap<&'a str, &'a str>);
@@ -32,12 +34,23 @@ impl<'a> Extractor for MetadataMap<'a> {
 ///
 /// TODO: Make Jaeger configurable
 ///
-pub fn init_jaeger_tracing(name: &str, endpoint: &str) -> Result<sdktrace::Tracer, TraceError> {
-    global::set_text_map_propagator(TraceContextPropagator::new());
-    opentelemetry_jaeger::new_agent_pipeline()
+pub fn init_jaeger_tracing(
+    name: String,
+    endpoint: &str,
+) -> Result<LoggerProvider, opentelemetry_sdk::logs::LogError> {
+    let exporter = LogExporter::builder()
+        .with_http()
         .with_endpoint(endpoint)
-        .with_service_name(name)
-        .install_simple()
+        .with_protocol(Protocol::HttpBinary)
+        .build()?;
+
+    Ok(LoggerProvider::builder()
+        .with_simple_exporter(exporter)
+        .with_resource(Resource::new([KeyValue::new(
+            Key::new("name"),
+            Value::String(name.into()),
+        )]))
+        .build())
 }
 
 pub fn serialize_context(context: &Context) -> String {