René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"`HttpServletStreamableServerTransportProvider` removes session on non-fatal failure","articleBody":"**Bug description**\nWe use the MCP Java SDK as via our dependency on Spring AI. Our CI/CD test coverage for our MCP implementation fails regularly with a race condition when dealing with intentional error conditions. We could not reproduce it locally on dev machines, but via logging eventually found a workaround to what appears to be a problem in HttpServletStreamableTransportProvider.\n\n**Environment**\nVersions 1.1.1, 1.1.2, and 2.0.0-M2 of the Java SDK. Prerelease versions of Spring AI 2.0, including 2.0.0-M5 and 2.0.0-M4. Java 25.\n\n**Steps to reproduce**\n\nSee attached for a standalone repro. The error manifested for us as a race condition but always of this form:\n\n```\n  java.lang.RuntimeException: MCP session with server terminated\n  at io.modelcontextprotocol.spec.McpClientSession.lambda$dismissPendingResponses$1(McpClientSession.java:128)\n  at java.base/java.util.concurrent.ConcurrentHashMap.forEach(ConcurrentHashMap.java:1619)\n  at io.modelcontextprotocol.spec.McpClientSession.dismissPendingResponses(McpClientSession.java:126)\n  at io.modelcontextprotocol.spec.McpClientSession.close(McpClientSession.java:304)\n  at io.modelcontextprotocol.client.LifecycleInitializer$DefaultInitialization.close(LifecycleInitializer.java:225)\n  at io.modelcontextprotocol.client.LifecycleInitializer.handleException(LifecycleInitializer.java:257)\n  at io.modelcontextprotocol.client.transport.HttpClientStreamableHttpTransport.handleException(HttpClientStreamableHttpTransport.java:231)\n  at io.modelcontextprotocol.client.transport.HttpClientStreamableHttpTransport.lambda$sendMessage$33(HttpClientStreamableHttpTransport.java:641)\n  at reactor.core.publisher.FluxOnErrorReturn$ReturnSubscriber.onError(FluxOnErrorReturn.java:172)\n  at reactor.core.publisher.FluxOnErrorResume$ResumeSubscriber.onError(FluxOnErrorResume.java:107)\n  at reactor.core.publisher.Operators.error(Operators.java:198)\n  at reactor.core.publisher.FluxError.subscribe(FluxError.java:44)\n  at reactor.core.publisher.Flux.subscribe(Flux.java:8888)\n  at reactor.core.publisher.FluxOnErrorResume$ResumeSubscriber.onError(FluxOnErrorResume.java:104)\n  at reactor.core.publisher.FluxFlatMap$FlatMapMain.checkTerminated(FluxFlatMap.java:847)\n  at reactor.core.publisher.FluxFlatMap$FlatMapMain.drainLoop(FluxFlatMap.java:613)\n  at reactor.core.publisher.FluxFlatMap$FlatMapMain.drain(FluxFlatMap.java:593)\n  at reactor.core.publisher.FluxFlatMap$FlatMapMain.onError(FluxFlatMap.java:456)\n  at reactor.core.publisher.SerializedSubscriber.onError(SerializedSubscriber.java:125)\n  at reactor.core.publisher.FluxRetryWhen$RetryWhenMainSubscriber.whenError(FluxRetryWhen.java:230)\n  at reactor.core.publisher.FluxRetryWhen$RetryWhenOtherSubscriber.onError(FluxRetryWhen.java:282)\n  at reactor.core.publisher.FluxFlatMap$FlatMapMain.checkTerminated(FluxFlatMap.java:847)\n  at reactor.core.publisher.FluxFlatMap$FlatMapMain.drainLoop(FluxFlatMap.java:613)\n  at reactor.core.publisher.FluxFlatMap$FlatMapMain.drain(FluxFlatMap.java:593)\n  at reactor.core.publisher.FluxFlatMap$FlatMapMain.onError(FluxFlatMap.java:456)\n  at reactor.core.publisher.FluxFlatMap$FlatMapMain.onNext(FluxFlatMap.java:419)\n  at reactor.core.publisher.SinkManyEmitterProcessor.drain(SinkManyEmitterProcessor.java:480)\n  at reactor.core.publisher.SinkManyEmitterProcessor.tryEmitNext(SinkManyEmitterProcessor.java:278)\n  at reactor.core.publisher.SinkManySerialized.tryEmitNext(SinkManySerialized.java:100)\n  at reactor.core.publisher.InternalManySink.emitNext(InternalManySink.java:27)\n  at reactor.core.publisher.FluxRetryWhen$RetryWhenMainSubscriber.onError(FluxRetryWhen.java:195)\n  at reactor.core.publisher.FluxFlatMap$FlatMapMain.checkTerminated(FluxFlatMap.java:847)\n  at reactor.core.publisher.FluxFlatMap$FlatMapMain.drainLoop(FluxFlatMap.java:613)\n  at reactor.core.publisher.FluxFlatMap$FlatMapMain.drain(FluxFlatMap.java:593)\n  at reactor.core.publisher.FluxFlatMap$FlatMapMain.onError(FluxFlatMap.java:456)\n  at reactor.core.publisher.FluxFlatMap$FlatMapMain.onNext(FluxFlatMap.java:419)\n  at reactor.core.publisher.MonoFlatMapMany$FlatMapManyInner.onNext(MonoFlatMapMany.java:251)\n  at reactor.core.publisher.FluxCreate$BufferAsyncSink.drain(FluxCreate.java:887)\n  at reactor.core.publisher.FluxCreate$BufferAsyncSink.next(FluxCreate.java:812)\n  at reactor.core.publisher.FluxCreate$SerializedFluxSink.next(FluxCreate.java:164)\n  at io.modelcontextprotocol.client.transport.ResponseSubscribers$AggregateSubscriber.hookOnComplete(ResponseSubscribers.java:263)\n  at reactor.core.publisher.BaseSubscriber.onComplete(BaseSubscriber.java:200)\n  at org.reactivestreams.FlowAdapters$FlowToReactiveSubscriber.onComplete(FlowAdapters.java:221)\n  at java.net.http/jdk.internal.net.http.LineSubscriberAdapter$LineSubscription.loop(LineSubscriberAdapter.java:430)\n  at java.net.http/jdk.internal.net.http.common.SequentialScheduler$LockingRestartableTask.run(SequentialScheduler.java:182)\n  at java.net.http/jdk.internal.net.http.common.SequentialScheduler$CompleteRestartableTask.run(SequentialScheduler.java:149)\n  at java.net.http/jdk.internal.net.http.common.SequentialScheduler$SchedulableTask.run(SequentialScheduler.java:207)\n  at java.net.http/jdk.internal.net.http.common.SequentialScheduler.runOrSchedule(SequentialScheduler.java:280)\n  at java.net.http/jdk.internal.net.http.common.SequentialScheduler.runOrSchedule(SequentialScheduler.java:233)\n  at java.net.http/jdk.internal.net.http.LineSubscriberAdapter$LineSubscription.signalComplete(LineSubscriberAdapter.java:193)\n  at java.net.http/jdk.internal.net.http.LineSubscriberAdapter.onComplete(LineSubscriberAdapter.java:114)\n  at java.net.http/jdk.internal.net.http.common.HttpBodySubscriberWrapper.complete(HttpBodySubscriberWrapper.java:293)\n  at java.net.http/jdk.internal.net.http.common.HttpBodySubscriberWrapper.onComplete(HttpBodySubscriberWrapper.java:401)\n  at java.net.http/jdk.internal.net.http.ResponseContent$ChunkedBodyParser.accept(ResponseContent.java:220)\n  at java.net.http/jdk.internal.net.http.ResponseContent$ChunkedBodyParser.accept(ResponseContent.java:131)\n  at java.net.http/jdk.internal.net.http.Http1Response$BodyReader.handle(Http1Response.java:708)\n  at java.net.http/jdk.internal.net.http.Http1Response$BodyReader.handle(Http1Response.java:636)\n  at java.net.http/jdk.internal.net.http.Http1Response$Receiver.accept(Http1Response.java:528)\n  at java.net.http/jdk.internal.net.http.Http1Response$BodyReader.tryAsyncReceive(Http1Response.java:666)\n  at java.net.http/jdk.internal.net.http.Http1AsyncReceiver.flush(Http1AsyncReceiver.java:233)\n  at java.net.http/jdk.internal.net.http.common.SequentialScheduler$LockingRestartableTask.run(SequentialScheduler.java:182)\n  at java.net.http/jdk.internal.net.http.common.SequentialScheduler$CompleteRestartableTask.run(SequentialScheduler.java:149)\n  at java.net.http/jdk.internal.net.http.common.SequentialScheduler$TryEndDeferredCompleter.complete(SequentialScheduler.java:324)\n  at java.net.http/jdk.internal.net.http.common.SequentialScheduler$CompleteRestartableTask.run(SequentialScheduler.java:151)\n  at java.net.http/jdk.internal.net.http.common.SequentialScheduler$SchedulableTask.run(SequentialScheduler.java:207)\n  at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1090)\n  at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:614)\n  at java.base/java.lang.Thread.run(Thread.java:1474)\n  Suppressed: java.lang.Exception: #block terminated with an error\n    at reactor.core.publisher.BlockingSingleSubscriber.blockingGet(BlockingSingleSubscriber.java:104)\n    at reactor.core.publisher.Mono.block(Mono.java:1773)\n    at io.modelcontextprotocol.client.McpSyncClient.callTool(McpSyncClient.java:236)\n    at org.labkey.professional.McpServerTest.callTool(McpServerTest.java:358)\n    at org.labkey.professional.McpServerTest.callToolExpectingFailure(McpServerTest.java:342)\n    at org.labkey.professional.McpServerTest.testBadTableParameters(McpServerTest.java:566)\n    at org.labkey.professional.McpServerTest.getSourceForSavedQuery(McpServerTest.java:522)\n    at org.labkey.professional.McpServerTest.invokeTools(McpServerTest.java:327)\n    at org.labkey.professional.McpServerTest.testAdminAccess(McpServerTest.java:254)\n    at java.base/jdk.internal.reflect.DirectMethodHandleAccessor.invoke(DirectMethodHandleAccessor.java:104)\n    at java.base/java.lang.reflect.Method.invoke(Method.java:565)\n    at org.junit.runners.model.FrameworkMethod$1.runReflectiveCall(FrameworkMethod.java:59)\n    at org.junit.internal.runners.model.ReflectiveCallable.run(ReflectiveCallable.java:12)\n```\n\n**Expected behavior**\nThe session should continue to process requests and respond to the client without fatal errors.\n\n**Minimal Complete Reproducible example**\nSee attached.\n\n[mcp-sdk-bug-repro.zip](https://github.com/user-attachments/files/27574476/mcp-sdk-bug-repro.zip)\n\n```bash\ncd mcp-sdk-bug-repro\nmvn -q test\n```\n\nThe test in `src/test/java/io/modelcontextprotocol/bug/SessionRemovalBugReproTest.java`:\n\n1. Builds a real `HttpServletStreamableServerTransportProvider` and a minimal `McpSyncServer` with a single tool.\n2. Issues an `initialize` POST and captures the server-assigned `mcp-session-id`.\n3. Issues a `tools/call` POST whose response writer is wired to a `Writer` that **accepts the writes** (so the SDK successfully serializes the SSE event into the response body) but **throws `IOException` on `flush()`** — exactly what an OutputStream backed by a closed socket does. This sets `PrintWriter.trouble = true` so `writer.checkError()` returns `true`.\n4. Asserts that:\n   - The full SSE event payload was written to the response (data was delivered).\n   - The session was removed from the SDK's `sessions` map (verified via reflection).\n5. Issues a second `tools/call` POST with the same session id and asserts that the SDK responds `404 Session not found` — proving the session is unrecoverable from the client's perspective even though the prior call's response was actually delivered.\n\nThe test deterministically reproduces what is otherwise a flaky race between a fast client closing its socket and the server's post-write `checkError()`.\n\n**Workaround**\nWrap the response in an `HttpServletResponseWrapper` that returns a `PrintWriter` whose `checkError()` always returns `false`. This suppresses the false positive at the cost of closing the (already broken) post-write disconnect detection. Apply only on the POST path — keep `checkError` honest for the long-lived GET listening stream.\n\n```\n    private static class _LoggingResponseWrapper extends HttpServletResponseWrapper\n    {\n        private final String requestUri;\n        private _LoggingPrintWriter loggingWriter;\n\n        _LoggingResponseWrapper(HttpServletResponse response, String requestUri)\n        {\n            super(response);\n            this.requestUri = requestUri;\n        }\n\n        @Override\n        public PrintWriter getWriter() throws IOException\n        {\n            if (loggingWriter == null)\n                loggingWriter = new _LoggingPrintWriter(super.getWriter(), requestUri);\n            return loggingWriter;\n        }\n    }\n\n    private static class _LoggingPrintWriter extends PrintWriter\n    {\n        private final String requestUri;\n\n        _LoggingPrintWriter(PrintWriter delegate, String requestUri)\n        {\n            super(delegate);\n            this.requestUri = requestUri;\n        }\n\n        @Override\n        public boolean checkError()\n        {\n            // Workaround for MCP SDK 1.1.2 bug in HttpServletStreamableServerTransportProvider.sendEvent\n            // (https://github.com/modelcontextprotocol/java-sdk). After writing/flushing an SSE event,\n            // sendEvent calls writer.checkError() and throws \"Client disconnected\" if it returns true.\n            // The catch block in sendMessage then removes the MCP session from the sessions map - even\n            // though the SSE event data was already delivered to the client. For streamable HTTP, each\n            // POST is a single-event exchange, so a client closing its socket after receiving the\n            // response is normal HTTP/1.1 behavior, not an MCP session termination. The race between\n            // the client's socket close and the server's checkError() call is exactly what makes this\n            // test flaky on TeamCity. Returning false suppresses the false positive; if the client\n            // truly lost data, it will retry on a fresh connection.\n            boolean actual = super.checkError();\n            if (actual)\n            {\n                LOG.info(\"MCP checkError suppressed (returning false): uri={}\", requestUri);\n            }\n            return false;\n        }\n    }\n```","author":{"url":"https://github.com/labkey-jeckels","@type":"Person","name":"labkey-jeckels"},"datePublished":"2026-05-10T21:43:59.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":2},"url":"https://github.com/952/java-sdk/issues/952"}