Reconcile the Frappe-specific gthread changes with upstream's major
26.0.0 release (HTTP/2, ASGI, lock-free main-thread event loop).
Conflict resolution:
- gunicorn/workers/gthread.py: reimplemented our features on top of
upstream's rewritten event loop instead of merging line-by-line:
- Per-request timeout + faulthandler traceback dump, tracked via an
in-flight future set and keyed on exec_start_time so client-wait
time (upstream's _DEFER path) is excluded and long-lived HTTP/2
connections are exempt.
- Adaptive slow/fast-lane queueing: the request-line peek/classify
now reuses upstream's pending_conns poller-park mechanism, then
dispatches to a fast or slow ThreadPoolExecutor.
- Dropped our eventfd-based shutdown wakeup in favour of upstream's
PollableMethodQueue.defer(), which already wakes the poller on
SIGTERM.
- gunicorn/config.py: kept both new validators; our EnableAdaptiveQueueing
and SlowRequestThreshold settings merged cleanly.
- README.rst was removed upstream (converted to README.md); ported an
updated Fork Information section into README.md.
- tox.ini / .github/workflows/tox.yml: took upstream's Python matrix
(it already drops the EoL versions our fork had pruned).
All gthread + routing tests pass (91), full suite 1976 passed / 263
skipped.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Eventlet was deprecated for 26.0 and is now removed:
- Delete gunicorn/workers/geventlet.py and its registry entry
- Drop eventlet from config help text, HTTP/2 unsupported-worker
messages, and the dirty client docstring
- Drop the eventlet optional-dependency, the eventlet entry in the
testing extra, and the eventlet-only filterwarnings ignore
- Drop the EventletWorkerAlpn test class
- Drop the freebsd CI ignore for the (now non-existent) test_geventlet.py
- Drop eventlet from the issue-triage discussion template
- Drop eventlet from README, install/design/http2/settings/news docs;
rewrite the news.md entry from 'deprecated' to 'removed in this release'
Add h2 and uvloop to requirements_test.txt so a plain
'pip install -r requirements_test.txt' run reaches feature parity with
'pip install .[testing]' for those two deps. The container suite
previously skipped 87 HTTP/2 tests for missing h2 and 1 for uvloop;
the in-process suite skips drop from 67 to 40.
- Remove Open Collective links from README, sponsor page, and FUNDING.yml
- Update Revolut donation link
- Add Enki Multimedia as sponsor on homepage, sponsor page, and README
The previous default /run/gunicorn.ctl requires root permissions.
Now uses $XDG_RUNTIME_DIR/gunicorn.ctl if available, otherwise
$HOME/.gunicorn/gunicorn.ctl. This works on Linux, FreeBSD, OpenBSD,
and macOS without requiring elevated privileges.
- Add _get_default_control_socket() helper in config.py
- Create parent directory automatically with 0o700 permissions
- Update gunicornc CLI to use the same default path
- Add unit tests for path selection and directory creation
Wire HttpParser to ASGI hot path, replacing AsyncRequest.parse() with
direct buffer-based parsing. Add FastAsyncRequest wrapper for body
reading. Replace per-request Queue/Task with BodyReceiver for on-demand
body reading. Keep headers as bytes end-to-end to avoid conversion
overhead. Add backpressure control and keepalive timer. Cache response
status lines and Date header.
Benchmark shows 3x improvement: ~875K req/s for simple GET (was ~340K).
- Read chunk size lines and trailers in 64-byte blocks instead of 1 byte
at a time, pushing back excess data to the unreader buffer
- Reuse BytesIO buffers with truncate/seek instead of creating new
objects to reduce GC pressure in AsyncUnreader
- Use bytearray.find() directly instead of converting to bytes first
in header parsing loop
- Use index-based iteration for header parsing instead of list.pop(0)
which is O(n) per pop vs O(1) for index access
Add tests for the optimized parsing code paths.
- Add guides/gunicornc.md with usage examples and command reference
- Update mkdocs.yml navigation to include Control Interface guide
- Update 2026-news.md and news.md changelog with 25.2.0 release
- Regenerate reference/settings.md with control socket settings
Add support for SIGTTIN and SIGTTOU signals to the dirty arbiter,
allowing dynamic scaling of dirty workers at runtime without restarting
gunicorn.
Changes:
- Add TTIN/TTOU to DirtyArbiter.SIGNALS
- Add num_workers instance variable for dynamic count
- Add _get_minimum_workers() to enforce app worker constraints
- Add signal handlers for TTIN (increase) and TTOU (decrease)
- Update manage_workers() to use dynamic count
- Add documentation for dynamic scaling
- Add unit tests for signal handling
- Add Docker integration tests
The minimum worker constraint ensures TTOU cannot reduce workers below
what apps require (e.g., if an app has workers=3, minimum is 3).
Closes#3489
New Features:
- Dirty Stash: global shared state between workers (#3503)
- Dirty Binary Protocol: TLV encoding for efficient IPC (#3500)
Documentation:
- Fix Markdown formatting in /configure
* feat(dirty): add stash - global shared state between workers
Add a simple key-value store (stash) that allows dirty workers to share
state through the arbiter. Tables are stored directly in arbiter memory
for fast access and simplicity.
Features:
- Auto-create tables on first access
- Dict-like interface via stash.table()
- Pattern matching for keys (glob patterns)
- Module-level API: stash.put(), stash.get(), stash.delete(), etc.
Usage:
from gunicorn.dirty import stash
stash.put("sessions", "user:1", {"name": "Alice"})
user = stash.get("sessions", "user:1")
# Or dict-like
sessions = stash.table("sessions")
sessions["user:1"] = {"name": "Alice"}
New files:
- gunicorn/dirty/stash.py - Client API and StashTable class
- Protocol additions for MSG_TYPE_STASH and STASH_OP_* codes
Note: Tables are ephemeral - lost if arbiter restarts.
* test(dirty): add tests for stash protocol and encoding
Test coverage for:
- Stash message creation and encoding
- Protocol constants (MSG_TYPE_STASH, STASH_OP_*)
- Error classes (StashError, StashTableNotFoundError, StashKeyNotFoundError)
- StashTable dict-like interface
- Edge cases: unicode, complex values, special patterns
* example(dirty): add stash usage example and integration tests
- Add SessionApp to dirty_app.py demonstrating stash usage
- Add /session/* endpoints to wsgi_app.py
- Add test_stash_integration.py with Docker tests
- Update docker-compose.yml with stash-test service
- Fix: Set GUNICORN_DIRTY_SOCKET in dirty arbiter for worker access
* docs(dirty): add stash documentation
- Update test_protocol.py example to use binary protocol
- Add test_binary_data_handling example showing raw bytes transfer
- Update dirty.md to document binary TLV protocol format
- Replace JSON references with binary protocol
- Add Binary Protocol section with header and TLV encoding details
- Add sponsor banner at top of README
- Add Revolut and Open Collective to FUNDING.yml
- Add dedicated sponsor page in docs
- Add sponsor heart link in footer
- Change title from "The Python WSGI Server" to "Serve Python Apps on the Web"
- Remove "Version {{ version }}" text from hero section
- Remove unused .hero__version CSS
- Bump version to 25.0.2
- Update copyright year to 2026 in LICENSE and NOTICE
- Add license headers to all Python source files
- Add changelog entry for 25.0.2
Closes#3484
When a client disconnects during an ASGI request, the worker now:
1. Sends http.disconnect message to the app's receive queue
2. Allows a configurable grace period for cleanup (default: 3 seconds)
3. Only cancels the task after the grace period expires
This follows the ASGI HTTP Connection Scope spec which defines
http.disconnect as the message apps should receive when clients
disconnect: https://asgi.readthedocs.io/en/latest/specs/www.html#disconnect-receive-event
The grace period prevents CancelledError from propagating to async
database operations, allowing SQLAlchemy and other async DB libraries
to properly reset their connection pools.
New config option: --asgi-disconnect-grace-period (default: 3 seconds)
