utils: explicit OTEL export config and OTEL enablement via common entry point (#11139)

We want to export performance traces from the pageserver in OTEL format.
End goal is to see them in Grafana.

To this end, there are two changes here:
1. Update the `tracing-utils` crate to allow for explicitly specifying
the export configuration. Pageserver configuration is loaded from a file
on start-up. This allows us to use the same flow for export configs
there.
2. Update the `utils::logging::init` common entry point to set up OTEL
tracing infrastructure if requested. Note that an entirely different
tracing subscriber is used. This is to avoid interference with the
existing tracing set-up. For now, no service uses this functionality.

PR to plug this into the pageserver is
[here](https://github.com/neondatabase/neon/pull/11140).

Related https://github.com/neondatabase/neon/issues/9873

Cargo.lock

@@ -4313,6 +4313,7 @@ dependencies = [
  "tokio-util",
  "toml_edit",
  "tracing",
+ "tracing-utils",
  "url",
  "utils",
  "uuid",
@@ -7850,6 +7851,7 @@ dependencies = [
  "tracing",
  "tracing-error",
  "tracing-subscriber",
+ "tracing-utils",
  "walkdir",
 ]
 

compute_tools/src/logger.rs

@@ -24,7 +24,8 @@ pub async fn init_tracing_and_logging(default_log_level: &str) -> anyhow::Result
         .with_writer(std::io::stderr);
 
     // Initialize OpenTelemetry
-    let otlp_layer = tracing_utils::init_tracing("compute_ctl").await;
+    let otlp_layer =
+        tracing_utils::init_tracing("compute_ctl", tracing_utils::ExportConfig::default()).await;
 
     // Put it all together
     tracing_subscriber::registry()

libs/tracing-utils/src/lib.rs

@@ -21,7 +21,7 @@
 //!         .with_writer(std::io::stderr);
 //!
 //!     // Initialize OpenTelemetry. Exports tracing spans as OpenTelemetry traces
-//!     let otlp_layer = tracing_utils::init_tracing("my_application").await;
+//!     let otlp_layer = tracing_utils::init_tracing("my_application", tracing_utils::ExportConfig::default()).await;
 //!
 //!     // Put it all together
 //!     tracing_subscriber::registry()
@@ -38,8 +38,12 @@ pub mod http;
 
 use opentelemetry::KeyValue;
 use opentelemetry::trace::TracerProvider;
-use tracing::Subscriber;
+use opentelemetry_otlp::WithExportConfig;
+pub use opentelemetry_otlp::{ExportConfig, Protocol};
+use tracing::level_filters::LevelFilter;
+use tracing::{Dispatch, Subscriber};
 use tracing_subscriber::Layer;
+use tracing_subscriber::layer::SubscriberExt;
 use tracing_subscriber::registry::LookupSpan;
 
 /// Set up OpenTelemetry exporter, using configuration from environment variables.
@@ -69,19 +73,28 @@ use tracing_subscriber::registry::LookupSpan;
 ///
 /// This doesn't block, but is marked as 'async' to hint that this must be called in
 /// asynchronous execution context.
-pub async fn init_tracing<S>(service_name: &str) -> Option<impl Layer<S>>
+pub async fn init_tracing<S>(
+    service_name: &str,
+    export_config: ExportConfig,
+) -> Option<impl Layer<S>>
 where
     S: Subscriber + for<'span> LookupSpan<'span>,
 {
     if std::env::var("OTEL_SDK_DISABLED") == Ok("true".to_string()) {
         return None;
     };
-    Some(init_tracing_internal(service_name.to_string()))
+    Some(init_tracing_internal(
+        service_name.to_string(),
+        export_config,
+    ))
 }
 
 /// Like `init_tracing`, but creates a separate tokio Runtime for the tracing
 /// tasks.
-pub fn init_tracing_without_runtime<S>(service_name: &str) -> Option<impl Layer<S>>
+pub fn init_tracing_without_runtime<S>(
+    service_name: &str,
+    export_config: ExportConfig,
+) -> Option<impl Layer<S>>
 where
     S: Subscriber + for<'span> LookupSpan<'span>,
 {
@@ -112,16 +125,22 @@ where
     ));
     let _guard = runtime.enter();
 
-    Some(init_tracing_internal(service_name.to_string()))
+    Some(init_tracing_internal(
+        service_name.to_string(),
+        export_config,
+    ))
 }
 
-fn init_tracing_internal<S>(service_name: String) -> impl Layer<S>
+fn init_tracing_internal<S>(service_name: String, export_config: ExportConfig) -> impl Layer<S>
 where
     S: Subscriber + for<'span> LookupSpan<'span>,
 {
-    // Sets up exporter from the OTEL_EXPORTER_* environment variables.
+    // Sets up exporter from the provided [`ExportConfig`] parameter.
+    // If the endpoint is not specified, it is loaded from the
+    // OTEL_EXPORTER_OTLP_ENDPOINT environment variable.
     let exporter = opentelemetry_otlp::SpanExporter::builder()
         .with_http()
+        .with_export_config(export_config)
         .build()
         .expect("could not initialize opentelemetry exporter");
 
@@ -151,3 +170,51 @@ where
 pub fn shutdown_tracing() {
     opentelemetry::global::shutdown_tracer_provider();
 }
+
+pub enum OtelEnablement {
+    Disabled,
+    Enabled {
+        service_name: String,
+        export_config: ExportConfig,
+        runtime: &'static tokio::runtime::Runtime,
+    },
+}
+
+pub struct OtelGuard {
+    pub dispatch: Dispatch,
+}
+
+impl Drop for OtelGuard {
+    fn drop(&mut self) {
+        shutdown_tracing();
+    }
+}
+
+/// Initializes OTEL infrastructure for performance tracing according to the provided configuration
+///
+/// Performance tracing is handled by a different [`tracing::Subscriber`]. This functions returns
+/// an [`OtelGuard`] containing a [`tracing::Dispatch`] associated with a newly created subscriber.
+/// Applications should use this dispatch for their performance traces.
+///
+/// The lifetime of the guard should match taht of the application. On drop, it tears down the
+/// OTEL infra.
+pub fn init_performance_tracing(otel_enablement: OtelEnablement) -> Option<OtelGuard> {
+    let otel_subscriber = match otel_enablement {
+        OtelEnablement::Disabled => None,
+        OtelEnablement::Enabled {
+            service_name,
+            export_config,
+            runtime,
+        } => {
+            let otel_layer = runtime
+                .block_on(init_tracing(&service_name, export_config))
+                .with_filter(LevelFilter::INFO);
+            let otel_subscriber = tracing_subscriber::registry().with(otel_layer);
+            let otel_dispatch = Dispatch::new(otel_subscriber);
+
+            Some(otel_dispatch)
+        }
+    };
+
+    otel_subscriber.map(|dispatch| OtelGuard { dispatch })
+}

libs/utils/Cargo.toml

@@ -42,6 +42,7 @@ toml_edit = { workspace = true, features = ["serde"] }
 tracing.workspace = true
 tracing-error.workspace = true
 tracing-subscriber = { workspace = true, features = ["json", "registry"] }
+tracing-utils.workspace = true
 rand.workspace = true
 scopeguard.workspace = true
 strum.workspace = true

libs/utils/src/logging.rs

@@ -165,6 +165,7 @@ pub fn init(
         };
         log_layer.with_filter(rust_log_env_filter())
     });
+
     let r = r.with(
         TracingEventCountLayer(&TRACING_EVENT_COUNT_METRIC).with_filter(rust_log_env_filter()),
     );

pageserver/Cargo.toml

@@ -70,6 +70,7 @@ tokio-stream.workspace = true
 tokio-util.workspace = true
 toml_edit = { workspace = true, features = [ "serde" ] }
 tracing.workspace = true
+tracing-utils.workspace = true
 url.workspace = true
 walkdir.workspace = true
 metrics.workspace = true

pageserver/compaction/tests/tests.rs

@@ -12,7 +12,7 @@ pub(crate) fn setup_logging() {
             logging::TracingErrorLayerEnablement::EnableWithRustLogFilter,
             logging::Output::Stdout,
         )
-        .expect("Failed to init test logging")
+        .expect("Failed to init test logging");
     });
 }
 

pageserver/src/bin/pageserver.rs

@@ -111,6 +111,7 @@ fn main() -> anyhow::Result<()> {
     } else {
         TracingErrorLayerEnablement::Disabled
     };
+
     logging::init(
         conf.log_format,
         tracing_error_layer_enablement,

pageserver/src/tenant.rs

@@ -5754,7 +5754,7 @@ pub(crate) mod harness {
                 logging::TracingErrorLayerEnablement::EnableWithRustLogFilter,
                 logging::Output::Stdout,
             )
-            .expect("Failed to init test logging")
+            .expect("Failed to init test logging");
         });
     }
 

proxy/src/logging.rs

@@ -46,7 +46,8 @@ pub async fn init() -> anyhow::Result<LoggingGuard> {
                 .expect("this should be a valid filter directive"),
         );
 
-    let otlp_layer = tracing_utils::init_tracing("proxy").await;
+    let otlp_layer =
+        tracing_utils::init_tracing("proxy", tracing_utils::ExportConfig::default()).await;
 
     let json_log_layer = if logfmt == LogFormat::Json {
         Some(JsonLoggingLayer::new(