StdioClientTransport starts and manages a child MCP server process. During graceful shutdown, the SDK transport currently calls process.destroy() and waits for process.onExit().
That works when the child process responds to SIGTERM, but it can hang indefinitely if the child process ignores termination or becomes stuck. Downstream, we had to implement a custom McpClientTransport to ensure proxied MCP server processes are always cleaned up when the parent shuts down.
Current behavior
The current shutdown flow is effectively:
if (this.process != null) {
this.process.destroy();
return Mono.fromFuture(process.onExit());
}
There is no configurable timeout and no fallback to destroyForcibly().
Expected behavior
StdioClientTransport.closeGracefully() should support bounded process termination:
- Stop accepting new messages.
- Complete the transport sinks.
- Send graceful termination with
process.destroy().
- Wait up to a configurable timeout.
- If the process is still alive, call
process.destroyForcibly().
- Dispose the transport schedulers.
This guarantees that child MCP server processes do not survive parent shutdown indefinitely.
Proposed fix
Add a configurable process termination timeout to StdioClientTransport, with a sensible default.
Conceptually:
process.destroy();
return Mono.fromFuture(process.onExit())
.timeout(processTerminationTimeout, Mono.defer(() -> {
if (process.isAlive()) {
process.destroyForcibly();
}
return Mono.fromFuture(process.onExit());
}));
This could be exposed through the existing ServerParameters builder or a dedicated StdioClientTransport constructor/builder option.
Suggested regression test
Add a test with a child process that ignores SIGTERM or does not exit promptly.
The test should verify that:
closeGracefully() completes within the configured timeout plus a small buffer.
- The child process is no longer alive after
closeGracefully() completes.
- The transport schedulers and sinks are cleaned up.
StdioClientTransportstarts and manages a child MCP server process. During graceful shutdown, the SDK transport currently callsprocess.destroy()and waits forprocess.onExit().That works when the child process responds to SIGTERM, but it can hang indefinitely if the child process ignores termination or becomes stuck. Downstream, we had to implement a custom
McpClientTransportto ensure proxied MCP server processes are always cleaned up when the parent shuts down.Current behavior
The current shutdown flow is effectively:
There is no configurable timeout and no fallback to
destroyForcibly().Expected behavior
StdioClientTransport.closeGracefully()should support bounded process termination:process.destroy().process.destroyForcibly().This guarantees that child MCP server processes do not survive parent shutdown indefinitely.
Proposed fix
Add a configurable process termination timeout to
StdioClientTransport, with a sensible default.Conceptually:
This could be exposed through the existing
ServerParametersbuilder or a dedicatedStdioClientTransportconstructor/builder option.Suggested regression test
Add a test with a child process that ignores SIGTERM or does not exit promptly.
The test should verify that:
closeGracefully()completes within the configured timeout plus a small buffer.closeGracefully()completes.