Remove max_requests feature from server

This feature (inherited from gunicorn) auto-restarted workers after N
requests to mitigate memory leaks. It was disabled by default and not
actively used. Removing it eliminates scattered conditional logic,
jitter calculations, and request counting from the worker and H2
handler hot paths.

Author: davegaeddert

plain/plain/cli/server.py

@@ -59,13 +59,6 @@
     setting="SERVER_ACCESS_LOG",
     help="Enable/disable access logging to stdout",
 )
-@click.option(
-    "--max-requests",
-    type=int,
-    cls=SettingOption,
-    setting="SERVER_MAX_REQUESTS",
-    help="Max requests before worker restart (0=disabled)",
-)
 @click.option(
     "--pidfile",
     type=click.Path(),
@@ -80,7 +73,6 @@ def server(
     keyfile: str | None,
     reload: bool,
     access_log: bool,
-    max_requests: int,
     pidfile: str | None,
 ) -> None:
     """Production-ready HTTP server"""
@@ -105,7 +97,6 @@ def server(
         threads=threads,
         workers=workers,
         timeout=timeout,
-        max_requests=max_requests,
         reload=reload,
         pidfile=pidfile,
         certfile=certfile,

plain/plain/runtime/global_settings.py

@@ -177,7 +177,6 @@
 SERVER_WORKERS: int = int(os.environ.get("WEB_CONCURRENCY", 0))  # 0 = auto (CPU count)
 SERVER_THREADS: int = 4
 SERVER_TIMEOUT: int = 30
-SERVER_MAX_REQUESTS: int = 0
 SERVER_ACCESS_LOG: bool = True
 SERVER_ACCESS_LOG_FIELDS: list[str] = [
     "method",

plain/plain/server/README.md

@@ -59,18 +59,17 @@ All options are available via the command line. Run `plain server --help` to see
 
 Most options can also be configured via settings (see below). CLI arguments take priority over settings.
 
-| Option                           | Setting               | Description                          |
-| -------------------------------- | --------------------- | ------------------------------------ |
-| `--bind` / `-b`                  | -                     | Address to bind (can repeat)         |
-| `--workers` / `-w`               | `SERVER_WORKERS`      | Worker processes (0=auto, CPU count) |
-| `--threads`                      | `SERVER_THREADS`      | Threads per worker                   |
-| `--timeout` / `-t`               | `SERVER_TIMEOUT`      | Worker timeout in seconds            |
-| `--access-log / --no-access-log` | `SERVER_ACCESS_LOG`   | Enable/disable access logging        |
-| `--max-requests`                 | `SERVER_MAX_REQUESTS` | Max requests before worker restart   |
-| `--reload`                       | -                     | Restart workers on code changes      |
-| `--certfile`                     | -                     | Path to SSL certificate file         |
-| `--keyfile`                      | -                     | Path to SSL key file                 |
-| `--pidfile`                      | -                     | PID file path                        |
+| Option                           | Setting             | Description                          |
+| -------------------------------- | ------------------- | ------------------------------------ |
+| `--bind` / `-b`                  | -                   | Address to bind (can repeat)         |
+| `--workers` / `-w`               | `SERVER_WORKERS`    | Worker processes (0=auto, CPU count) |
+| `--threads`                      | `SERVER_THREADS`    | Threads per worker                   |
+| `--timeout` / `-t`               | `SERVER_TIMEOUT`    | Worker timeout in seconds            |
+| `--access-log / --no-access-log` | `SERVER_ACCESS_LOG` | Enable/disable access logging        |
+| `--reload`                       | -                   | Restart workers on code changes      |
+| `--certfile`                     | -                   | Path to SSL certificate file         |
+| `--keyfile`                      | -                   | Path to SSL key file                 |
+| `--pidfile`                      | -                   | PID file path                        |
 
 ## Settings
 
@@ -80,7 +79,6 @@ Server behavior can be configured in your `settings.py` file. These are the defa
 SERVER_WORKERS = 0           # 0 = auto (one per CPU core)
 SERVER_THREADS = 4
 SERVER_TIMEOUT = 30
-SERVER_MAX_REQUESTS = 0      # 0 = disabled
 SERVER_ACCESS_LOG = True
 SERVER_ACCESS_LOG_FIELDS = ["method", "path", "query", "status", "duration_ms", "size", "ip", "user_agent", "referer"]
 SERVER_GRACEFUL_TIMEOUT = 30

plain/plain/server/app.py

@@ -24,7 +24,6 @@ def __init__(
         workers: int,
         threads: int,
         timeout: int,
-        max_requests: int,
         reload: bool,
         pidfile: str | None,
         certfile: str | None,
@@ -35,7 +34,6 @@ def __init__(
         self.workers = workers
         self.threads = threads
         self.timeout = timeout
-        self.max_requests = max_requests
         self.reload = reload
         self.pidfile = pidfile
         self.certfile = certfile

plain/plain/server/http/h2handler.py

@@ -297,15 +297,12 @@ async def async_handle_h2_connection(
     handler: Any,
     is_ssl: bool,
     executor: ThreadPoolExecutor,
-    max_requests: int = 0,
-) -> int:
+) -> None:
     """Async HTTP/2 connection loop.
 
     Reads frames from the socket in a dedicated thread (to avoid exhausting
     the shared executor pool), dispatches each completed stream as an
     independent asyncio task.
-
-    Returns the number of streams (requests) completed.
     """
     config = h2.config.H2Configuration(client_side=False)
     conn = h2.connection.H2Connection(config=config)
@@ -324,7 +321,6 @@ async def async_handle_h2_connection(
     sock.settimeout(30)
 
     stream_tasks: dict[int, asyncio.Task[None]] = {}
-    streams_completed = 0
 
     recv_queue: asyncio.Queue[bytes | None] = asyncio.Queue()
     reader_stop = threading.Event()
@@ -432,10 +428,8 @@ def _reader_thread() -> None:
                         def _on_stream_done(
                             t: asyncio.Task[None], sid: int = event.stream_id
                         ) -> None:
-                            nonlocal streams_completed
                             stream_tasks.pop(sid, None)
                             state.cleanup_stream(sid)
-                            streams_completed += 1
 
                         task.add_done_callback(_on_stream_done)
 
@@ -467,12 +461,6 @@ def _on_stream_done(
             else:
                 async with state.write_lock:
                     await state.flush()
-                if max_requests and streams_completed >= max_requests:
-                    log.info(
-                        "HTTP/2 connection reached max_requests (%d), closing",
-                        max_requests,
-                    )
-                    break
                 continue
             # break from for-loop reached — exit while-loop too
             break
@@ -508,8 +496,6 @@ def _on_stream_done(
         if reader_thread.is_alive():
             log.warning("H2 reader thread did not exit cleanly")
 
-    return streams_completed
-
 
 async def _async_handle_stream(
     state: H2ConnectionState,

plain/plain/server/workers/thread.py

@@ -26,7 +26,6 @@
 import time
 from concurrent.futures import ThreadPoolExecutor
 from datetime import datetime
-from random import randint
 from types import FrameType
 from typing import TYPE_CHECKING, Any
 
@@ -63,9 +62,6 @@ class _H2Sentinel:
 
 _H2_SENTINEL = _H2Sentinel()
 
-# Maximum jitter to add to max_requests to stagger worker restarts
-MAX_REQUESTS_JITTER = 50
-
 # Keep-alive connection timeout in seconds
 KEEPALIVE = 2
 
@@ -159,14 +155,6 @@ def __init__(
         self.booted = False
         self.reloader: Any = None
 
-        self.nr = 0
-
-        if app.max_requests > 0:
-            jitter = randint(0, MAX_REQUESTS_JITTER)
-            self.max_requests = app.max_requests + jitter
-        else:
-            self.max_requests = sys.maxsize
-
         self.alive = True
         self.log = logging.getLogger("plain.server")
         self.heartbeat = heartbeat
@@ -313,27 +301,15 @@ async def _handle_connection(self, conn: TConn) -> None:
                 ) = await loop.run_in_executor(self.tpool, self._parse_request, conn)
 
                 if isinstance(parse_result, _H2Sentinel):
-                    remaining = self.max_requests - self.nr
-                    if self.max_requests < sys.maxsize and remaining <= 0:
-                        self.log.info("Autorestarting worker after current request.")
-                        self.alive = False
-                        return
                     conn.handed_off = True
-                    h2_streams = await async_handle_h2_connection(
+                    await async_handle_h2_connection(
                         conn.sock,
                         conn.client,
                         conn.server,
                         self.handler,
                         self.app.is_ssl,
                         self.tpool,
-                        max_requests=remaining
-                        if self.max_requests < sys.maxsize
-                        else 0,
                     )
-                    self.nr += h2_streams
-                    if self.nr >= self.max_requests:
-                        self.log.info("Autorestarting worker after current request.")
-                        self.alive = False
                     return
 
                 if parse_result is None:
@@ -571,12 +547,6 @@ def _parse_request(
             http_request = create_request(req, conn.sock, conn.client, conn.server)
 
             resp = Response(req, conn.sock, is_ssl=self.app.is_ssl)
-            self.nr += 1
-            if self.nr >= self.max_requests:
-                if self.alive:
-                    self.log.info("Autorestarting worker after current request.")
-                    self.alive = False
-                resp.force_close()
 
             if not self.alive:
                 resp.force_close()