tbench2 env server: enable idle-session reaper to reclaim leaked slots

[Shi-Dong]

What

The tbench2_env server now enables the core idle-session reaper by passing an explicit ConcurrencyConfig(max_concurrent_envs=..., session_timeout=...) to create_app, instead of the bare max_concurrent_envs= argument. A new SESSION_IDLE_TIMEOUT_S env var (default 3600) controls the reap interval.

Why

The /mcp WebSocket handler releases its session slot and Docker container only in a finally block, which runs solely on a clean socket close. Half-open or uncleanly-dropped connections never raise WebSocketDisconnect, so receive_text() blocks forever, the finally never runs, and the session slot + its Docker container leak.

Under sustained load these leaks accumulate over roughly a day until all MAX_CONCURRENT_ENVS slots are permanently held. New episodes then deadlock on CAPACITY_REACHED and the only recovery is a manual env-server restart.

Passing a bare max_concurrent_envs maps to ConcurrencyConfig(session_timeout=None) in the core server, which disables _reap_idle_sessions. Supplying an explicit session_timeout re-enables the reaper so stale sessions are evicted and their containers reclaimed automatically, without operator intervention.

Notes

• Default SESSION_IDLE_TIMEOUT_S=3600 preserves existing behavior for short-lived sessions while bounding leaked slots.
• Single-file change scoped to envs/tbench2_env/server/app.py; no change to the core server or to the on-the-wire protocol.

---

Note

Low Risk
Single-file server bootstrap change with a conservative default timeout; no protocol or core server edits.

Overview
The tbench2_env server now passes an explicit ConcurrencyConfig to create_app (with max_concurrent_envs and session_timeout) instead of only max_concurrent_envs, which turns on the core idle-session reaper that was previously off when session_timeout was unset.

A new SESSION_IDLE_TIMEOUT_S env var (default 3600) sets how long a session can sit idle before it is reaped so its WebSocket slot and Docker container are released. That targets leaks from half-open /mcp WebSocket connections that never hit a clean disconnect, which could otherwise hold all MAX_CONCURRENT_ENVS slots until a manual restart.

^{Reviewed by Cursor Bugbot for commit adc5d44. Bugbot is set up for automated code reviews on this repo. Configure here.}

[bot-ci-comment]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[nblintao]

LGTM.

[caiotheodoro]

src/openenv/core/env_server/http_server.py:1168 — risk: this is the first env in the repo to turn the idle reaper on (grep shows no other env passes session_timeout), and it surfaces a gap in the shared reaper path rather than one in this diff. mcp_websocket_endpoint closes over session_env as a local and never rereads self._sessions. _destroy_session (line 441) pops the session and calls env.close() — for tbench2 that stops+removes the Docker container — without touching the websocket itself. A client that's slow rather than dead, past the 1hr SESSION_IDLE_TIMEOUT_S, gets its container killed mid-connection; its next message still runs mcp_handler against the now-closed session_env and comes back as an opaque INTERNAL_ERROR instead of a clean timeout signal. Worth closing the websocket (or flagging the session so the handler bails cleanly) from _destroy_session before this ships, since it's the first PR to make that path reachable.

[cursor]

Stale comment

Still not merge-ready despite the earlier human approval on adc5d447.

• Branch is 117 commits behind main (mergeStateStatus: BEHIND). Refresh and re-run exact-head CI before any merge attempt.
• The follow-up comment from caiotheodoro is a real product concern: enabling the idle reaper for the first time can destroy a live Docker-backed session while the WebSocket is still open, turning a slow client into an opaque INTERNAL_ERROR. Prefer closing/signaling the websocket from session destroy, or otherwise fail cleanly, before this becomes the first production caller of that path.
• Automation cannot update the PR branch (403). Author/maintainer refresh needed.

  

_{Sent by Cursor Automation: Release}

[cursor]

Stale comment

Refresh review vs main b12450f2 (Ben ask):

Verdict: LGTM after rebase. Exact head adc5d447 is ~129 behind main; historical CI was green and nblintao APPROVED. Diff correctly enables core idle reaper via ConcurrencyConfig(session_timeout=SESSION_IDLE_TIMEOUT_S) (default 3600) for tbench2 capacity leaks from half-open sockets. Env-local, low risk. Please rebase onto current main and re-run CI, then this is squash-merge ready.

View PR

  

_{Sent by Cursor Automation: Release}

[burtenshaw]

Thanks

[cursor]

Refresh review (Wed 2026-09-16)

Still BEHIND main (adc5d447). Diff is still only enabling session_timeout in envs/tbench2_env/server/app.py.

Prior blocker stands: core _destroy_session does not close/signal the MCP WebSocket. Enabling the reaper makes tbench2 the first production caller of that broken path → opaque INTERNAL_ERROR / hung receive for slow clients. #1169’s /ws?session_id= attach exemptions do not protect ordinary /mcp WS sessions.

Also: please remove the copy of this change that was folded into #1172 until the core signaling fix exists.

Needs author/core fix — not mergeable after refresh alone.

View PR

  

_{Sent by Cursor Automation: Release}