From c22fcd49b551875919b087ab9dc88dc6a94a72fd Mon Sep 17 00:00:00 2001
From: Manuel Fedele <manuel.fedele@generali.com>
Date: Mon, 16 Mar 2026 09:31:28 +0100
Subject: [PATCH] fix(chroma): close transport before nulling to prevent
 subprocess leak

In callTool(), both the initial catch and retry catch blocks set
this.transport = null without calling transport.close() first.
When connectInternal() runs next, it sees transport === null,
skips the cleanup, and spawns a new chroma-mcp subprocess while
the old one is still alive. Over time this leaks dozens of orphaned
uvx/chroma-mcp processes, each downloading a 79MB ONNX model,
exhausting RAM and causing MCP search timeouts.

Fix: save a reference to the stale transport, null the instance
field, then close the saved reference. This ensures the subprocess
receives SIGTERM before a replacement is spawned.

Fixes #1369
---
 src/services/sync/ChromaMcpManager.ts | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/src/services/sync/ChromaMcpManager.ts b/src/services/sync/ChromaMcpManager.ts
index abd027655..73f6af98c 100644
--- a/src/services/sync/ChromaMcpManager.ts
+++ b/src/services/sync/ChromaMcpManager.ts
@@ -252,9 +252,17 @@ export class ChromaMcpManager {
       // Transport error: chroma-mcp subprocess likely died (e.g., killed by orphan reaper,
       // HNSW index corruption). Mark connection dead and retry once after reconnect (#1131).
       // Without this retry, callers see a one-shot error even though reconnect would succeed.
+      //
+      // CRITICAL: Close the transport before nulling to kill the subprocess (#1369).
+      // If we null this.transport first, connectInternal() sees transport === null,
+      // skips cleanup, and spawns a new subprocess, orphaning the old one.
       this.connected = false;
+      const staleTransport = this.transport;
       this.client = null;
       this.transport = null;
+      if (staleTransport) {
+        try { await staleTransport.close(); } catch { /* subprocess may already be dead */ }
+      }
 
       logger.warn('CHROMA_MCP', `Transport error during "${toolName}", reconnecting and retrying once`, {
         error: transportError instanceof Error ? transportError.message : String(transportError)
@@ -267,7 +275,14 @@ export class ChromaMcpManager {
           arguments: toolArguments
         });
       } catch (retryError) {
+        // Retry also failed. Close transport to prevent another orphan.
         this.connected = false;
+        const retryTransport = this.transport;
+        this.client = null;
+        this.transport = null;
+        if (retryTransport) {
+          try { await retryTransport.close(); } catch { /* subprocess may already be dead */ }
+        }
         throw new Error(`chroma-mcp transport error during "${toolName}" (retry failed): ${retryError instanceof Error ? retryError.message : String(retryError)}`);
       }
     }