From cce5b66dba401ce6bef3c624f5ff17343a8cb849 Mon Sep 17 00:00:00 2001
From: Liona Can <liona@MacBook-Air-de-Liona.local>
Date: Thu, 13 Aug 2026 05:44:40 -0400
Subject: [PATCH] fix(observability): shut down tracer provider

---
 src/bin/dolos/common.rs | 19 +++++++++++++++++++
 src/bin/dolos/main.rs   |  7 +++++--
 2 files changed, 24 insertions(+), 2 deletions(-)

diff --git a/src/bin/dolos/common.rs b/src/bin/dolos/common.rs
index 8e74c9bf4..9cf5b83c3 100644
--- a/src/bin/dolos/common.rs
+++ b/src/bin/dolos/common.rs
@@ -7,7 +7,9 @@ use dolos_snapshot::registry::Auth;
 use miette::{Context as _, IntoDiagnostic};
 use opentelemetry::trace::TracerProvider as _;
 use opentelemetry_otlp::WithExportConfig as _;
+use opentelemetry_sdk::trace::SdkTracerProvider;
 use std::sync::Arc;
+use std::sync::OnceLock;
 use std::{fs, path::PathBuf, time::Duration};
 use tokio_util::sync::CancellationToken;
 use tracing::{debug, info};
@@ -20,6 +22,8 @@ use dolos::storage;
 
 pub type Stores = storage::Stores<dolos_cardano::CardanoDelta>;
 
+static TRACER_PROVIDER: OnceLock<SdkTracerProvider> = OnceLock::new();
+
 /// Ensure the storage root directory exists.
 pub fn ensure_storage_path(config: &RootConfig) -> Result<PathBuf, Error> {
     storage::ensure_storage_path(config)
@@ -285,6 +289,7 @@ pub fn setup_tracing(config: &LoggingConfig, telemetry: &TelemetryConfig) -> mie
             .build();
 
         opentelemetry::global::set_tracer_provider(tracer.clone());
+        let _ = TRACER_PROVIDER.set(tracer.clone());
 
         let layer = tracing_opentelemetry::layer().with_tracer(tracer.tracer("dolos"));
         Some(layer)
@@ -319,6 +324,20 @@ pub fn setup_tracing(config: &LoggingConfig, telemetry: &TelemetryConfig) -> mie
     Ok(())
 }
 
+/// Flush and stop the batch span exporter before the process exits.
+///
+/// The global provider owns a clone, so merely dropping the local provider from
+/// `setup_tracing` does not stop its worker or export its final batch. Keeping a
+/// handle here lets the CLI perform an explicit shutdown after every command,
+/// including commands that do not run the long-lived daemon pipeline.
+pub fn shutdown_tracing() {
+    if let Some(provider) = TRACER_PROVIDER.get() {
+        if let Err(error) = provider.shutdown() {
+            eprintln!("failed to shut down OpenTelemetry tracer provider: {error}");
+        }
+    }
+}
+
 pub fn open_genesis_files(config: &GenesisConfig) -> miette::Result<Genesis> {
     Genesis::from_file_paths(
         &config.byron_path,
diff --git a/src/bin/dolos/main.rs b/src/bin/dolos/main.rs
index a10aeb429..3625269aa 100644
--- a/src/bin/dolos/main.rs
+++ b/src/bin/dolos/main.rs
@@ -92,7 +92,7 @@ fn main() -> Result<()> {
 
     let feedback = crate::feedback::Feedback::default();
 
-    match (config, args.command) {
+    let result = match (config, args.command) {
         (Ok(config), Command::Daemon(args)) => daemon::run(config, &args),
         (Ok(config), Command::Sync(args)) => sync::run(&config, &args),
         (Ok(config), Command::Serve(args)) => serve::run(config, &args),
@@ -119,5 +119,8 @@ fn main() -> Result<()> {
         (Ok(config), Command::Minikupo(x)) => minikupo::run(&config, &x),
 
         (Err(x), _) => Err(x),
-    }
+    };
+
+    crate::common::shutdown_tracing();
+    result
 }