SSE: ConnectionClosed exception after session disconnect

[tomconte]

Initial Checks

• I confirm that I'm using the latest version of MCP Python SDK
• I confirm that I searched for my issue in https://github.com/modelcontextprotocol/python-sdk/issues before opening this issue

Description

In our environment, the SSE connections sometimes get dropped:

DEBUG:sse_starlette.sse:Got event: http.disconnect. Stop streaming.
DEBUG:root:Client session disconnected 081769ba-c683-471b-af22-1333409ab669

When this happens, on the next request for that session, we get a ClosedResourceError exception:

INFO: 100.100.0.154:49316 - "POST /messages/?session_id=081769bac683471baf221333409ab669 HTTP/1.1" 202 Accepted
ERROR:__main__:SSE post message error:
Traceback (most recent call last):
File "/app/mcp_renewables/server.py", line 714, in handle_post_message_with_error_handling
return await sse.handle_post_message(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/app/.venv/lib/python3.12/site-packages/mcp/server/sse.py", line 249, in handle_post_message
await writer. Send(session_message)
File "/app/.venv/lib/python3.12/site-packages/anyio/streams/memory.py", line 242, in send
self.send_nowait(item)
File "/app/.venv/lib/python3.12/site-packages/anyio/streams/memory.py", line 211, in send_nowait
raise ClosedResourceError
anyio.ClosedResourceError

I believe the reason is that when the session disconnects, its file handles are not properly removed as explained in #703. I have forked the repo and applied the fix and it seems to solve the problem.

Example Code: the fix I applied

diff --git a/src/mcp/server/sse.py b/src/mcp/server/sse.py
index b7ff332..29acecb 100644
--- a/src/mcp/server/sse.py
+++ b/src/mcp/server/sse.py
@@ -179,7 +179,7 @@ class SseServerTransport:
 
         async with anyio.create_task_group() as tg:
 
-            async def response_wrapper(scope: Scope, receive: Receive, send: Send):
+            async def response_wrapper(scope: Scope, receive: Receive, send: Send, transport: SseServerTransport):
                 """
                 The EventSourceResponse returning signals a client close / disconnect.
                 In this case we close our side of the streams to signal the client that
@@ -190,10 +190,13 @@ class SseServerTransport:
                 )
                 await read_stream_writer.aclose()
                 await write_stream_reader.aclose()
+                await read_stream.aclose()
+                await write_stream.aclose()
+                transport._read_stream_writers.pop(session_id)
                 logging.debug(f"Client session disconnected {session_id}")
 
             logger.debug("Starting SSE response task")
-            tg.start_soon(response_wrapper, scope, receive, send)
+            tg.start_soon(response_wrapper, scope, receive, send, self)
 
             logger.debug("Yielding read and write streams")
             yield (read_stream, write_stream)

Python & MCP Python SDK

1.12.2

References

Related PRs:

• Handle SSE Disconnects Properly #612
• Fix clean up sse disconnect event (not upgrade lib) #703

[felixweinberger]

Hi @tomconte thanks for submitting this issue - could you provide a code snippet or example I could run to reproduce the error? Then I could debug further on my machine.