openclaw - 💡(How to fix) Fix MCP Server Memory Leak - zombie processes accumulate [1 comments, 2 participants]

TL;DR

The most likely fix is to modify the Gateway's MCP tool handler to properly reap forked child processes.

Guidance

• Investigate the Gateway's MCP tool handler code to identify the issue with reaping child processes.
• Consider using a mechanism like waitpid or a process group to manage and clean up child processes.
• Verify the fix by running the reproduction steps and checking for zombie processes using ps aux | grep defunct.
• To mitigate the issue temporarily, use the provided pkill -f mcp-tfs-query command to clean up accumulated processes.

Example

# Example of using waitpid to reap a child process
pid=$!
wait $pid

Notes

The provided temporary fix using pkill may not be suitable for production use, as it may terminate processes unexpectedly. A proper fix should be implemented in the Gateway's MCP tool handler.

Recommendation

Apply a workaround by modifying the Gateway's MCP tool handler to properly reap child processes, as the root cause is identified and a temporary fix is available.