aio-libs/aiohttp · error · RuntimeError

Using chunked encoding is forbidden for HTTP/{request.versio

Error message

Using chunked encoding is forbidden for HTTP/{request.version.major}.{request.version.minor}

What it means

Raised during response preparation when chunked transfer-encoding is enabled (`_chunked = True`) but the request was made over HTTP/1.0 (or any version other than 1.1). Chunked encoding is defined only in HTTP/1.1 (RFC 9112); emitting it on a 1.0 connection would corrupt the response. aiohttp refuses rather than sending malformed data.

Source

Thrown at aiohttp/web_response.py:393

        writer = self._payload_writer
        assert writer is not None
        keep_alive = self._keep_alive
        if keep_alive is None:
            keep_alive = request.keep_alive
        self._keep_alive = keep_alive

        version = request.version

        headers = self._headers
        if self._cookies:
            populate_with_cookies(headers, self._cookies)

        if self._compression:
            await self._start_compression(request)

        if self._chunked:
            if version != HttpVersion11:
                raise RuntimeError(
                    "Using chunked encoding is forbidden "
                    f"for HTTP/{request.version.major}.{request.version.minor}"
                )
            if not self._must_be_empty_body:
                writer.enable_chunking()
                headers[hdrs.TRANSFER_ENCODING] = "chunked"
        elif self._length_check:  # Disabled for WebSockets
            writer.length = self.content_length
            if writer.length is None:
                if version >= HttpVersion11:
                    if not self._must_be_empty_body:
                        writer.enable_chunking()
                        headers[hdrs.TRANSFER_ENCODING] = "chunked"
                elif not self._must_be_empty_body:
                    keep_alive = False

        # HTTP 1.1: https://tools.ietf.org/html/rfc7230#section-3.3.2
        # HTTP 1.0: https://tools.ietf.org/html/rfc1945#section-10.4

View on GitHub (pinned to d041d4d0fd)

Solutions

1. Return a response with an explicit Content-Length so chunking is not needed (buffer the body fully).
2. Ensure upstream clients/proxies speak HTTP/1.1; configure your reverse proxy to forward 1.1.
3. Close the connection instead of streaming: set `resp.content_length` indirectly by using a fully-buffered body so no chunking is required.
4. If you control the client, stop passing `-0` / `--http1.0` to curl or upgrade the client library.

Example fix

// before (streaming, no content-length, HTTP/1.0 client)
resp = StreamResponse(status=200)
resp.content_length = None
await resp.prepare(request)
await resp.write(chunk)  # forces chunked -> RuntimeError on 1.0

// after
body = b''.join(chunks)
return Response(body=body, status=200)  # Content-Length set automatically

def stream_or_buffer(request, body_chunks):
    if request.version != HttpVersion11:
        # HTTP/1.0 cannot chunk — buffer fully
        return Response(body=b''.join(body_chunks))
    resp = StreamResponse()
    resp.content_length = None
    return resp

from aiohttp.http import HttpVersion11

def supports_chunked(request) -> bool:
    return request.version == HttpVersion11

When it happens

Trigger: Client connects via HTTP/1.0 (e.g. `curl -0`, some legacy proxies, or Python's httplib in HTTP/1.0 mode) and the handler returns a streaming Response with no Content-Length, forcing chunked encoding. Also reproducible by enabling `enable_chunking()` on a writer for a 1.0 request, or forcing `Transfer-Encoding: chunked` on a response served to a 1.0 client.

Common situations: A proxy/load-balancer downgrades the protocol to HTTP/1.0; curl with `--http1.0` for testing; an embedded/legacy HTTP client speaking 1.0; misconfigured reverse proxy stripping Content-Length and forcing chunked upstream while advertising 1.0 downstream.

Understand the failure class

• Authentication and authorization failures — expired tokens, bad credentials, and missing scopes.

Related errors

• Cannot call write() after write_eof()
• Cannot call write() before prepare()
• Setting charset for application/octet-stream doesn't make se
• Unsupported type for last_modified: {type(value).__name__}
• Unsupported etag type: {type(value)}. etag must be str, ETag

AI-assisted analysis of aio-libs/aiohttp@d041d4d0fd (2026-08-11). Data as JSON: /api/errors/0a065c1d8f02ed62. Report an issue: GitHub.