# _drain URL: /pounce/api/_drain/ Section: api Description: Shared drain-503 wire format and write helpers. Single source of truth for the HTTP 503 response that every worker mode emits to *new* connections while the server is draining (SIGTERM / SIGHUP). Keeping the bytes in one place means the async ``Worker``, the blocking ``SyncWorker`` and the ``AcceptDistributor`` answer byte-identically: HTTP/1.1 503 Service Unavailable Connection: close Retry-After: 1 Content-Type: text/plain; charset=utf-8 Content-Length: 23 Server shutting down... The body is short and actionable; ``Retry-After`` tells well-behaved clients (and load balancers) to come back, while ``Connection: close`` guarantees the socket is torn down after the single response. --- > For a complete page index, fetch /pounce/llms.txt. Open LLM text (/pounce/api/_drain/index.txt) Share with AI Ask Claude (https://claude.ai/new?q=Please%20help%20me%20understand%20this%20documentation%3A%20%2Fpounce%2Fapi%2F_drain%2Findex.txt) Ask ChatGPT (https://chatgpt.com/?q=Please%20help%20me%20understand%20this%20documentation%3A%20%2Fpounce%2Fapi%2F_drain%2Findex.txt) Ask Gemini (https://gemini.google.com/app?q=Please%20help%20me%20understand%20this%20documentation%3A%20%2Fpounce%2Fapi%2F_drain%2Findex.txt) Ask Copilot (https://copilot.microsoft.com/?q=Please%20help%20me%20understand%20this%20documentation%3A%20%2Fpounce%2Fapi%2F_drain%2Findex.txt) Module # `_drain` Shared drain-503 wire format and write helpers. Single source of truth for the HTTP 503 response that every worker mode emits to new connections while the server is draining (SIGTERM / SIGHUP). Keeping the bytes in one place means the async`Worker`, the blocking `SyncWorker` and the `AcceptDistributor`answer byte-identically: ``` HTTP/1.1 503 Service Unavailable Connection: close Retry-After: 1 Content-Type: text/plain; charset=utf-8 Content-Length: 23 Server shutting down... ``` The body is short and actionable;`Retry-After`tells well-behaved clients (and load balancers) to come back, while`Connection: close` guarantees the socket is torn down after the single response. 8Functions ## Functions `_draining_health_method` 2 `bytes | None` ▼ Return GET/HEAD when *request_head* targets the configured health path. `def _draining_health_method(request_head: bytes, health_check_path: str | None) -> bytes | None` ##### Parameters Name Type Description `request_head` `bytes` `health_check_path` `str | None` ##### Returns `bytes | None` `_draining_health_response` 3 `bytes` ▼ Serialize the configured readiness response for the drain boundary. `def _draining_health_response(method: bytes, *, worker_id: int, active_connections: int) -> bytes` ##### Parameters Name Type Description `method` `bytes` `worker_id` `int` `active_connections` `int` ##### Returns `bytes` `_read_request_head_sync` 3 `bytes` ▼ Read one bounded HTTP/1 header block from a draining socket. `def _read_request_head_sync(conn: socket.socket, config: ServerConfig, *, timeout: float | None) -> bytes` ##### Parameters Name Type Description `conn` `socket.socket` `config` `ServerConfig (/pounce/api/config/#ServerConfig)` `timeout` `float | None` ##### Returns `bytes` `_read_request_head_async` 3 `bytes` ▼ Read one bounded HTTP/1 header block from a draining stream. async `async def _read_request_head_async(reader: asyncio.StreamReader, config: ServerConfig, *, timeout: float | None) -> bytes` ##### Parameters Name Type Description `reader` `asyncio.StreamReader` `config` `ServerConfig (/pounce/api/config/#ServerConfig)` `timeout` `float | None` ##### Returns `bytes` `write_drain_503_sync` 1 `None` ▼ Send the drain 503 on a blocking socket; the caller still owns the close. Tole… `def write_drain_503_sync(conn: socket.socket) -> None` Send the drain 503 on a blocking socket; the caller still owns the close. Tolerates a client that has already gone away — a dropped connection during shutdown is expected, not exceptional. ##### Parameters Name Type Description `conn` `socket.socket` `write_drain_response_sync` 5 `None` ▼ Inspect one bounded request and write the matching drain response. `def write_drain_response_sync(conn: socket.socket, config: ServerConfig, *, worker_id: int, active_connections: int, timeout: float | None = None) -> None` ##### Parameters Name Type Description `conn` `socket.socket` `config` `ServerConfig (/pounce/api/config/#ServerConfig)` `worker_id` `int` `active_connections` `int` `timeout` `float | None` Default:`None` `write_drain_503_async` 2 `None` ▼ Send the drain 503 on an asyncio writer and close it. Tolerates a client that … async `async def write_drain_503_async(writer: asyncio.StreamWriter, *, timeout: float = 30.0) -> None` Send the drain 503 on an asyncio writer and close it. Tolerates a client that has already gone away — a dropped connection during shutdown is expected, not exceptional. ##### Parameters Name Type Description `writer` `asyncio.StreamWriter` `timeout` `float` Default:`30.0` `write_drain_response_async` 6 `None` ▼ Inspect one bounded request, write its drain response, and close. async `async def write_drain_response_async(reader: asyncio.StreamReader, writer: asyncio.StreamWriter, config: ServerConfig, *, worker_id: int, active_connections: int, timeout: float | None = None) -> None` ##### Parameters Name Type Description `reader` `asyncio.StreamReader` `writer` `asyncio.StreamWriter` `config` `ServerConfig (/pounce/api/config/#ServerConfig)` `worker_id` `int` `active_connections` `int` `timeout` `float | None` Default:`None`