memory: clean up active lock files on exit and termination signals #5485

Which server?

memory (@modelcontextprotocol/server-memory)

Server version

2.1.0

How did you run the server?

From a clone of this repository

Transport

stdio

Protocol era

Modern (2026-07-28)

MCP client and version

Claude Desktop / Autonomous Agent Orchestrator

Operating system and runtime

Linux / Node.js v24+

Server configuration

{
  "mcpServers": {
    "memory": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-memory"]
    }
  }
}

Steps to reproduce

1. Run @modelcontextprotocol/server-memory.
2. Dispatch a mutation tool call (e.g. create_entities or add_observations) that acquires the lock via withFileLock.
3. Terminate or stop the server process via SIGINT, SIGTERM, client disconnect, or container stop during the operation.
4. Inspect the directory holding memory.jsonl.
5. Attempt a subsequent write from another agent or container sharing the same graph file.

Expected behavior

The server should register signal handlers (SIGINT, SIGTERM) and exit cleanup hooks to remove any active lock file it currently holds, preventing orphaned lock files on abrupt shutdown.

Actual behavior

In Node.js, asynchronous finally blocks in withFileLock do not run when terminated by signals. Because src/memory registers no signal handlers (unlike src/everything/transports/stdio.ts), memory.jsonl.lock remains stranded on disk.
In multi-container or Docker environments where hostname !== os.hostname(), isLockOwnerDead returns false, forcing other agents to stall for up to 30–60 seconds before breaking the stale lock.
Additionally, in src/memory/Dockerfile, Node runs as PID 1 without STOPSIGNAL SIGINT, causing PID 1 to ignore SIGINT by default.

Already prototyped a fix?

Prompt for maintainer workflow:

"In src/memory/index.ts, track active lock files held by this process in an activeLocks map. On process exit and on SIGINT/SIGTERM, synchronously check if the lock file still contains our token, and if so, unlink it to avoid leaving stranded locks on abrupt shutdown. Add STOPSIGNAL SIGINT to src/memory/Dockerfile matching src/everything/Dockerfile. Add unit tests in file-lock.test.ts verifying active lock tracking and that cleanup leaves locks owned by another process intact."

Verification:

• Implemented and verified: all 231 tests in src/memory passing.
• Working branch pushed for reference: https://github.com/Faris-as/servers/tree/memory-lock-signal-cleanup

Before you submit

• I searched existing issues and this is not a duplicate.
• This is not a security vulnerability report.