openai/codex
1
0
Fork 0
mirror of https://github.com/openai/codex.git synced 2026-10-31 19:28:49 +00:00
Code Issues Packages Projects Releases Wiki Activity

Preserve agent state when unloading idle multi-agent v2 children (#52034)

Browse Source 
## What changed

- Route app-server idle unloading of multi-agent v2 children through local eviction that preserves unread mail and environment selections for reload.
- Hold residency pins through input submission so newly reloaded children cannot be evicted before receiving their work.
- Guard listener cleanup against replacement runtimes and subscriptions, and reject local eviction of host-controlled threads.

## Testing

Add coverage for unsubscribe-driven eviction followed by `followup_task`, verifying restored conversation context and exactly-once delivery of queued mail. Extend tests to check that residency pins block eviction and that rejected host eviction preserves the thread and mailbox.

GitOrigin-RevId: e36723730af683d2081e495e637f0e6d23548b8d
This commit is contained in:
jif
2026-10-08 09:39:33 +00:00
committed by copyberry
parent ea27864f99
commit d2a6dac20b
12 changed files with 493 additions and 124 deletions
Download Patch File Download Diff File Expand all files Collapse all files
codex-rs/core/src/agent/control/spawn.rs
+39 -30
View File
@@ -320,12 +320,13 @@ impl LocalAgentControl {
}
/// A provided parent enables owner-validated reloads; `None` preserves sender-driven reloads.
/// Keep the returned residency pin until the caller has submitted its input.
pub(crate) async fn ensure_v2_agent_loaded(
&self,
mut config: Config,
thread_id: ThreadId,
parent: Option<Arc<CodexThread>>,
) -> CodexResult<()> {
) -> CodexResult<Option<tokio::sync::OwnedRwLockReadGuard<()>>> {
let membership = self.runtime.admit_start()?;
let state = self.runtime.upgrade()?;
let owner_thread_id = parent.as_ref().map(|parent| parent.session.thread_id);
@@ -347,9 +348,10 @@ impl LocalAgentControl {
)));
}
}
if owner_thread_id.is_none() && state.get_thread(thread_id).await.is_ok() {
self.touch_loaded_v2_residency(&state, thread_id).await;
return Ok(());
if owner_thread_id.is_none()
&& let Ok(thread) = state.get_thread(thread_id).await
{
return self.runtime.pin_v2_residency(&state, &thread).await;
}
if self
.runtime
@@ -402,8 +404,7 @@ impl LocalAgentControl {
}
if let Ok(thread) = state.get_thread(thread_id).await {
self.validate_loaded_v2_child(&thread, parent_thread_id)?;
self.touch_loaded_v2_residency(&state, thread_id).await;
return Ok(());
return self.runtime.pin_v2_residency(&state, &thread).await;
}
}
let parent = if let Some(parent) = parent {
@@ -633,28 +634,37 @@ impl LocalAgentControl {
if let Some(parent_thread_id) = owner_thread_id {
self.validate_loaded_v2_child(&reloaded_thread.thread, parent_thread_id)?;
}
self.runtime.registry.clear_evicted_environments(thread_id);
residency_slot.commit(reloaded_thread.thread_id);
// Register before listeners can forward events from the resumed thread.
if let Some((client_metadata, source)) = subagent_analytics {
let thread_config = reloaded_thread.thread.config_snapshot().await;
emit_subagent_session_started(
&reloaded_thread
.thread
.session
.services
.analytics_events_client,
client_metadata,
reloaded_thread.thread.session.session_id(),
reloaded_thread.thread_id,
thread_config.parent_thread_id,
thread_config,
source,
/*resumed_created_at*/ Some(stored_thread.created_at),
);
}
state.notify_thread_created(reloaded_thread.thread_id);
Ok(())
let runtime = self.runtime.clone();
// Finish registration even if the sender is cancelled while acquiring its pin.
tokio::spawn(async move {
let _membership = membership;
let residency_pin = runtime
.pin_v2_residency(&state, &reloaded_thread.thread)
.await?;
runtime.registry.clear_evicted_environments(thread_id);
residency_slot.commit(reloaded_thread.thread_id);
// Register before listeners can forward events from the resumed thread.
if let Some((client_metadata, source)) = subagent_analytics {
let thread_config = reloaded_thread.thread.config_snapshot().await;
emit_subagent_session_started(
&reloaded_thread
.thread
.session
.services
.analytics_events_client,
client_metadata,
reloaded_thread.thread.session.session_id(),
reloaded_thread.thread_id,
thread_config.parent_thread_id,
thread_config,
source,
/*resumed_created_at*/ Some(stored_thread.created_at),
);
}
state.notify_thread_created(reloaded_thread.thread_id);
Ok(residency_pin)
})
.await?
}
Err(err) => {
if let Ok(thread) = state.get_thread(thread_id).await {
@@ -663,8 +673,7 @@ impl LocalAgentControl {
}
self.runtime.registry.clear_evicted_environments(thread_id);
drop(residency_slot);
self.touch_loaded_v2_residency(&state, thread_id).await;
return Ok(());
return self.runtime.pin_v2_residency(&state, &thread).await;
}
Err(err)
}