Skip to content

Run each async udf in its own thread#124

Open
nkitsaini wants to merge 15 commits into
mainfrom
users/asaini/event-loop-worker
Open

Run each async udf in its own thread#124
nkitsaini wants to merge 15 commits into
mainfrom
users/asaini/event-loop-worker

Conversation

@nkitsaini

@nkitsaini nkitsaini commented Jun 3, 2026

Copy link
Copy Markdown
Collaborator

Note

High Risk
Changes core UDF concurrency, cancellation, and embeddings HTTP behavior under load; misconfiguration or regressions could affect timeouts, pooling, and concurrent EMBED_TEXT-style workloads.

Overview
Reworks Python UDF / external-function execution so async calls run on a single dedicated asyncio dispatch thread (fixing loop-bound resource reuse for httpx pools and enabling concurrent async requests), while sync UDFs use persistent per-worker-thread event loops instead of asyncio.run per request. Adds _cancellable_run with threading.Event cancellation, offloads heavy parse/serialize to the executor, and tightens cancel signaling on timeout/disconnect.

Embeddings get a default httpx.AsyncClient with env-tunable timeouts, keep-alive pooling, optional transport IP pin, and TracingAsyncTransport plus http_trace_id for correlated HTTP stage logging. Managed UDF apps set uvicorn timeout_keep_alive=120 to reduce upstream proxy connect timeouts.

Bumps package version to 1.16.11rc1+byasaini and adds test_udf_event_loop.py covering dispatch, cancellation, and Application routing.

Reviewed by Cursor Bugbot for commit 018db7a. Bugbot is set up for automated code reviews on this repo. Configure here.

Comment thread singlestoredb/functions/ext/asgi.py
Comment thread .github/workflows/publish.yml Outdated
Comment thread singlestoredb/functions/ext/asgi.py
@nkitsaini nkitsaini force-pushed the users/asaini/event-loop-worker branch from 2f24109 to 53e6c9a Compare June 3, 2026 17:01
@nkitsaini nkitsaini force-pushed the users/asaini/event-loop-worker branch from 53e6c9a to 78a96f8 Compare June 3, 2026 18:20
Comment thread .github/workflows/publish.yml Outdated
kesmit13 and others added 11 commits June 9, 2026 20:22
Async UDFs were running directly in uvicorn's event loop via
asyncio.create_task, competing with connection handling under heavy
concurrent load. This caused unresponsiveness when running from Jupyter
notebooks where the event loop is shared.

The fix introduces a dedicated event loop in a background thread for
async UDF execution. Coroutines are submitted via
run_coroutine_threadsafe() and awaited from the server loop, isolating
UDF work from HTTP I/O while preserving cooperative async scheduling
between UDFs.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Cancel the concurrent.futures.Future in the UDF loop on
disconnect/timeout so the coroutine is interrupted promptly,
not just at the next cancel_on_event row check.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…ture

asyncio.create_task() requires a coroutine but asyncio.wrap_future()
returns a Future. Use asyncio.ensure_future() which accepts both.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Prevents UDF event loop thread leaks when run_udf_app() is called
repeatedly in Jupyter notebooks.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Cancel udf_future when func_task is in pending set after asyncio.wait
- Cancel udf_future in finally block to ensure cleanup on any exit path
- Wrap post-construction code in try/except to call app.shutdown() if
  validation, config, or registration fails after Application is created

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Move udf_future initialization before input_handler['load']() to
  prevent NameError in finally block if parsing raises
- Lazily create UDF event loop on first async UDF invocation instead
  of unconditionally in __init__, avoiding wasted resources for
  sync-only or metadata-only usage
- Guard shutdown() against None loop/thread

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
After stopping the event loop and joining the thread, set both
_udf_loop and _udf_thread back to None so that _get_udf_loop()
can safely recreate them if called after shutdown.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The dedicated shared event loop still caused starvation under concurrent
async UDF calls. Switch to the same model used by sync UDFs: each request
gets its own thread with asyncio.run(), eliminating loop contention.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Wraps the inner coroutine in _cancellable_run which polls cancel_event
and raises CancelledError at the next await (~100ms), ensuring vector
UDFs respect disconnect/timeout signals without waiting for completion.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace asyncio.run() with _run_with_graceful_shutdown() that drains
pending callbacks before closing the loop, preventing RuntimeError from
httpx/anyio TLS cleanup in async UDFs calling OpenAI/LangChain APIs.

Add 17 unit tests covering graceful shutdown, cancellation timing,
exception propagation, context variable isolation, and concurrent safety.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
@nkitsaini nkitsaini force-pushed the users/asaini/event-loop-worker branch from 4c8958b to f27488e Compare June 9, 2026 14:53

@cursor cursor Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Cursor Bugbot has reviewed your changes using default effort and found 2 potential issues.

Fix All in Cursor

❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.

Reviewed by Cursor Bugbot for commit 018db7a. Configure here.

if cancel_check in done:
task.cancel()
raise asyncio.CancelledError()
return task.result()

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Cancel races successful UDF

High Severity

In _cancellable_run, when cancel_event is set at the same time the UDF coroutine finishes, asyncio.wait can place both the work task and _poll_cancel in done. The handler always treats a finished cancel poll as cancellation and raises CancelledError, even if the UDF task already completed successfully, so timeouts/disconnects can win over a valid result.

Fix in Cursor Fix in Web

Reviewed by Cursor Bugbot for commit 018db7a. Configure here.

'SINGLESTOREDB_EMBEDDINGS_POOL_TIMEOUT', '10',
),
),
)

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sync timeouts ignored async client

Medium Severity

SingleStoreEmbeddingsFactory parses connect_timeout and read_timeout from the optional sync http_client for Bedrock, but the new default httpx.AsyncClient for OpenAI embeddings only uses env defaults (60s read). Callers who configure timeouts via http_client get different limits on async embedding HTTP than they expect.

Fix in Cursor Fix in Web

Reviewed by Cursor Bugbot for commit 018db7a. Configure here.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants