[hyperactor] telemetry / mesh: display execution ID and a scuba shorturl for the execution

Pull Request resolved: #1909

This way we don't have to grub around logs to get to the execution id and corresponding trace. With this change, we provide a direct URL that you can click on in your terminal.

Differential Revision: [D87273699](https://our.internmc.facebook.com/intern/diff/D87273699/)
ghstack-source-id: 324277125

Author: mariusae

hyperactor_mesh/src/v1/host_mesh.rs

@@ -1416,10 +1416,27 @@ mod tests {
             .spawn(&instance, "test", Extent::unity())
             .await
             .unwrap_err();
-        assert_matches!(
-            err, v1::Error::ProcCreationError { state: resource::State { status: resource::Status::Failed(msg), ..}, .. }
-            if msg.contains("failed to configure process: Terminal(Stopped { exit_code: 1")
-        );
+        eprintln!("error: {}", err);
+        if let v1::Error::ProcCreationError {
+            state:
+                resource::State {
+                    status: resource::Status::Failed(msg),
+                    ..
+                },
+            ..
+        } = err
+        {
+            assert_eq!(
+                msg,
+                "failed to configure process: Terminal(Stopped { exit_code: 1"
+            );
+        } else {
+            panic!();
+        }
+        // assert_matches!(
+        //     err, v1::Error::ProcCreationError { state: resource::State { status: resource::Status::Failed(msg), ..}, .. }
+        //     if msg.contains("failed to configure process: Terminal(Stopped { exit_code: 1")
+        // );
     }
 
     #[tokio::test]

hyperactor_mesh/src/v1/host_mesh/mesh_agent.rs

@@ -132,8 +132,14 @@ impl Actor for HostMeshAgent {
             )
             .unwrap();
             eprintln!(
-                "Monarch internal logs are being written to {}/{}.log",
-                directory, file
+                "Monarch internal logs are being written to {}/{}.log; execution id {}{}",
+                directory,
+                file,
+                hyperactor_telemetry::env::execution_id(),
+                hyperactor_telemetry::env::execution_url()
+                    .await
+                    .unwrap_or_else(|e| Some(format!(": <error generating URL: {}>", e)))
+                    .map_or_else(|| "".to_string(), |url| format!(": {}", url))
             );
         }
         Ok(Self {

hyperactor_telemetry/Cargo.toml

@@ -32,6 +32,7 @@ tracing-appender = "0.2.3"
 tracing-core = { version = "0.1.33", features = ["valuable"] }
 tracing-glog = { version = "0.4.1", features = ["ansi", "tracing-log"] }
 tracing-subscriber = { version = "0.3.20", features = ["chrono", "env-filter", "json", "local-time", "parking_lot", "registry"] }
+urlencoding = "2.1.0"
 whoami = "1.5"
 
 [dev-dependencies]

hyperactor_telemetry/src/lib.rs

@@ -12,6 +12,7 @@
 #![feature(mpmc_channel)]
 #![feature(cfg_version)]
 #![feature(formatting_options)]
+#![recursion_limit = "256"]
 
 // TODO:ehedeman Remove or replace with better config once telemetry perf issues are solved
 /// Environment variable to disable the OpenTelemetry logging layer.
@@ -721,6 +722,24 @@ pub mod env {
         id
     }
 
+    /// Returns a URL for the execution trace, if available.
+    #[cfg(fbcode_build)]
+    pub async fn execution_url() -> anyhow::Result<Option<String>> {
+        let fb = if fbinit::was_performed() {
+            fbinit::expect_init()
+        } else {
+            // Safety: This is going to be embedded in a python library, so we can't be sure when fbinit has been called.
+            unsafe { fbinit::perform_init() }
+        };
+        Ok(Some(
+            crate::meta::scuba_tracing::url::get_samples_shorturl(fb, &execution_id()).await?,
+        ))
+    }
+    #[cfg(not(fbcode_build))]
+    pub async fn execution_url() -> anyhow::Result<Option<String>> {
+        Ok(None)
+    }
+
     #[derive(PartialEq)]
     pub enum Env {
         Local,