feat(runtime): add runtime-owned observability endpoints

crates/sof-observer/README.md

@@ -115,11 +115,19 @@ use std::net::SocketAddr;
 async fn main() -> Result<(), sof::runtime::RuntimeError> {
     let setup = sof::runtime::RuntimeSetup::new()
         .with_bind_addr(SocketAddr::from(([0, 0, 0, 0], 8001)))
+        .with_observability_bind_addr(SocketAddr::from(([127, 0, 0, 1], 9108)))
         .with_startup_step_logs(true);
     sof::runtime::run_async_with_setup(&setup).await
 }
 ```
 
+When `SOF_OBSERVABILITY_BIND` (or `RuntimeSetup::with_observability_bind_addr`) is set, the
+packaged runtime also serves:
+
+- `/metrics`
+- `/healthz`
+- `/readyz`
+
 Or apply one typed gossip/ingest profile instead of stringly env overrides:
 
 ```rust

crates/sof-observer/src/app/config/mod.rs

@@ -1,11 +1,13 @@
 mod base;
 mod common;
 mod gossip;
+mod observability;
 mod relay;
 mod repair;
 
 pub use base::*;
 pub use common::*;
 pub use gossip::*;
+pub use observability::*;
 pub use relay::*;
 pub use repair::*;

crates/sof-observer/src/app/config/observability.rs

@@ -0,0 +1,9 @@
+use std::net::SocketAddr;
+
+use super::read_env_var;
+
+/// Reads the runtime-owned observability bind address from `SOF_OBSERVABILITY_BIND`.
+#[must_use]
+pub fn read_observability_bind_addr() -> Option<SocketAddr> {
+    read_env_var("SOF_OBSERVABILITY_BIND").and_then(|value| value.parse::<SocketAddr>().ok())
+}

crates/sof-observer/src/app/runtime/entrypoints.rs

@@ -178,14 +178,40 @@ async fn run_async_with_hosts_and_optional_shutdown(
     derived_state_host: DerivedStateHost,
     shutdown_signal: Option<ShutdownSignal>,
 ) -> Result<(), RuntimeEntrypointError> {
-    runloop::run_async_with_hosts(
+    let observability = if let Some(bind_addr) = read_observability_bind_addr() {
+        let service = RuntimeObservabilityService::start(
+            bind_addr,
+            extension_host.clone(),
+            derived_state_host.clone(),
+        )
+        .await
+        .map_err(|source| RuntimeEntrypointError::Runloop {
+            reason: format!("failed to start observability endpoint on {bind_addr}: {source}"),
+        })?;
+        tracing::info!(
+            bind_addr = %service.local_addr(),
+            "runtime observability endpoint enabled"
+        );
+        Some(service)
+    } else {
+        None
+    };
+    let observability_handle = observability
+        .as_ref()
+        .map(RuntimeObservabilityService::handle)
+        .cloned();
+    let result = runloop::run_async_with_hosts(
         plugin_host,
         extension_host,
         derived_state_host,
         shutdown_signal,
+        observability_handle,
     )
-    .await
-    .map_err(|source| RuntimeEntrypointError::Runloop {
+    .await;
+    if let Some(service) = observability {
+        service.shutdown().await;
+    }
+    result.map_err(|source| RuntimeEntrypointError::Runloop {
         reason: source.to_string(),
     })
 }
@@ -257,15 +283,41 @@ pub(crate) async fn run_async_with_hosts_and_kernel_bypass_ingress(
     shutdown_signal: Option<ShutdownSignal>,
     packet_ingest_rx: ingest::RawPacketBatchReceiver,
 ) -> Result<(), RuntimeEntrypointError> {
-    runloop::run_async_with_hosts_and_kernel_bypass_ingress(
+    let observability = if let Some(bind_addr) = read_observability_bind_addr() {
+        let service = RuntimeObservabilityService::start(
+            bind_addr,
+            extension_host.clone(),
+            derived_state_host.clone(),
+        )
+        .await
+        .map_err(|source| RuntimeEntrypointError::Runloop {
+            reason: format!("failed to start observability endpoint on {bind_addr}: {source}"),
+        })?;
+        tracing::info!(
+            bind_addr = %service.local_addr(),
+            "runtime observability endpoint enabled"
+        );
+        Some(service)
+    } else {
+        None
+    };
+    let observability_handle = observability
+        .as_ref()
+        .map(RuntimeObservabilityService::handle)
+        .cloned();
+    let result = runloop::run_async_with_hosts_and_kernel_bypass_ingress(
         plugin_host,
         extension_host,
         derived_state_host,
         shutdown_signal,
+        observability_handle,
         packet_ingest_rx,
     )
-    .await
-    .map_err(|source| RuntimeEntrypointError::Runloop {
+    .await;
+    if let Some(service) = observability {
+        service.shutdown().await;
+    }
+    result.map_err(|source| RuntimeEntrypointError::Runloop {
         reason: source.to_string(),
     })
 }

crates/sof-observer/src/app/runtime/mod.rs

@@ -2,6 +2,7 @@ mod bootstrap;
 mod dataset;
 mod entrypoints;
 mod logging;
+mod observability;
 mod prelude;
 mod runloop;
 #[cfg(test)]
@@ -20,6 +21,7 @@ pub(crate) use entrypoints::{
     run_async_with_plugin_host_and_kernel_bypass_ingress,
 };
 use logging::init_tracing;
+use observability::{RuntimeObservabilityHandle, RuntimeObservabilityService};
 use prelude::*;
 
 #[cfg(feature = "gossip-bootstrap")]