feat(app-server, core): add more spans (#14479)

## Description

This PR expands tracing coverage across app-server thread startup, core
session initialization, and the Responses transport layer. It also gives
core dispatch spans stable operation-specific names so traces are easier
to follow than the old generic `submission_dispatch` spans.

Also use `fmt::Display` for types that we serialize in traces so we send
strings instead of rust types
This commit is contained in:
Owen Lin
2026-03-13 13:16:33 -07:00
committed by GitHub
parent 914f7c7317
commit 014e19510d
17 changed files with 473 additions and 88 deletions
+62 -12
View File
@@ -584,7 +584,6 @@ impl Codex {
let session_source_clone = session_configuration.session_source.clone();
let (agent_status_tx, agent_status_rx) = watch::channel(AgentStatus::PendingInit);
let session_init_span = info_span!("session_init");
let session = Session::new(
session_configuration,
config.clone(),
@@ -601,7 +600,6 @@ impl Codex {
file_watcher,
agent_control,
)
.instrument(session_init_span)
.await
.map_err(|e| {
error!("Failed to create session: {e:#}");
@@ -1340,6 +1338,7 @@ impl Session {
}
}
#[instrument(name = "session_init", level = "info", skip_all)]
#[allow(clippy::too_many_arguments)]
async fn new(
mut session_configuration: SessionConfiguration,
@@ -1431,18 +1430,29 @@ impl Session {
.await?;
Ok((Some(rollout_recorder), state_db_ctx))
}
};
}
.instrument(info_span!(
"session_init.rollout",
otel.name = "session_init.rollout",
session_init.ephemeral = config.ephemeral,
));
let is_subagent = matches!(
session_configuration.session_source,
SessionSource::SubAgent(_)
);
let history_meta_fut = async {
if matches!(
session_configuration.session_source,
SessionSource::SubAgent(_)
) {
if is_subagent {
(0, 0)
} else {
crate::message_history::history_metadata(&config).await
}
};
}
.instrument(info_span!(
"session_init.history_metadata",
otel.name = "session_init.history_metadata",
session_init.is_subagent = is_subagent,
));
let auth_manager_clone = Arc::clone(&auth_manager);
let config_for_mcp = Arc::clone(&config);
let mcp_manager_for_mcp = Arc::clone(&mcp_manager);
@@ -1455,7 +1465,11 @@ impl Session {
)
.await;
(auth, mcp_servers, auth_statuses)
};
}
.instrument(info_span!(
"session_init.auth_mcp",
otel.name = "session_init.auth_mcp",
));
// Join all independent futures.
let (
@@ -1613,7 +1627,12 @@ impl Session {
tx
};
let thread_name =
match session_index::find_thread_name_by_id(&config.codex_home, &conversation_id).await
match session_index::find_thread_name_by_id(&config.codex_home, &conversation_id)
.instrument(info_span!(
"session_init.thread_name_lookup",
otel.name = "session_init.thread_name_lookup",
))
.await
{
Ok(name) => name,
Err(err) => {
@@ -1663,6 +1682,12 @@ impl Session {
managed_network_requirements_enabled,
network_proxy_audit_metadata,
)
.instrument(info_span!(
"session_init.network_proxy",
otel.name = "session_init.network_proxy",
session_init.managed_network_requirements_enabled =
managed_network_requirements_enabled,
))
.await?;
(Some(network_proxy), Some(session_network_proxy))
} else {
@@ -1812,6 +1837,8 @@ impl Session {
.map(|(name, _)| name.clone())
.collect();
required_mcp_servers.sort();
let enabled_mcp_server_count = mcp_servers.values().filter(|server| server.enabled).count();
let required_mcp_server_count = required_mcp_servers.len();
let tool_plugin_provenance = mcp_manager.tool_plugin_provenance(config.as_ref());
{
let mut cancel_guard = sess.services.mcp_startup_cancellation_token.lock().await;
@@ -1829,6 +1856,12 @@ impl Session {
codex_apps_tools_cache_key(auth),
tool_plugin_provenance,
)
.instrument(info_span!(
"session_init.mcp_manager_init",
otel.name = "session_init.mcp_manager_init",
session_init.enabled_mcp_server_count = enabled_mcp_server_count,
session_init.required_mcp_server_count = required_mcp_server_count,
))
.await;
{
let mut manager_guard = sess.services.mcp_connection_manager.write().await;
@@ -1848,6 +1881,11 @@ impl Session {
.read()
.await
.required_startup_failures(&required_mcp_servers)
.instrument(info_span!(
"session_init.required_mcp_wait",
otel.name = "session_init.required_mcp_wait",
session_init.required_mcp_server_count = required_mcp_server_count,
))
.await;
if !failures.is_empty() {
let details = failures
@@ -4269,11 +4307,23 @@ async fn submission_loop(sess: Arc<Session>, config: Arc<Config>, rx_sub: Receiv
}
fn submission_dispatch_span(sub: &Submission) -> tracing::Span {
let op_name = sub.op.kind();
let span_name = format!("op.dispatch.{op_name}");
let dispatch_span = match &sub.op {
Op::RealtimeConversationAudio(_) => {
debug_span!("submission_dispatch", submission.id = sub.id.as_str())
debug_span!(
"submission_dispatch",
otel.name = span_name.as_str(),
submission.id = sub.id.as_str(),
codex.op = op_name
)
}
_ => info_span!("submission_dispatch", submission.id = sub.id.as_str()),
_ => info_span!(
"submission_dispatch",
otel.name = span_name.as_str(),
submission.id = sub.id.as_str(),
codex.op = op_name
),
};
if let Some(trace) = sub.trace.as_ref()
&& !set_parent_from_w3c_trace_context(&dispatch_span, trace)