openclaw - 💡(How to fix) Fix Gateway memory growth due to repeated file read errors

Bug Description

Gateway memory grows steadily (approx. 60MB/day) due to repeated [tools] read failed errors. These errors occur every 8 hours, coinciding with cron task execution.

Environment

• OpenClaw version: 2026.4.11 (769908e)
• OS: Ubuntu 24.04.4 LTS
• Node.js: v24.14.0
• Installation: npm global

Error Patterns

Three types of read errors occur regularly:

1. Wrong year in path (e.g., 2025-05-27 instead of 2026-05-27)

   [tools] read failed: ENOENT: no such file or directory, access '/home/dell/.openclaw/workspace/memory/2025-05-27.md'

   • Occurs at: 00:22, 04:22, 12:22, 18:22 (every 8 hours)

2. Wrong directory for CORE.md

   [tools] read failed: ENOENT: no such file or directory, access '/home/dell/.openclaw/workspace/memory-system/CORE.md'

   • Actual location: /home/dell/.openclaw/workspace/CORE.md
   • Occurs at: 01:35, 09:35, 12:47, 20:47

3. Attempting to read directory as file

   [tools] read failed: EISDIR: illegal operation on a directory, read
   path: '/home/dell/.openclaw/workspace/memory/agent-tasks'

   • Occurs at: 03:52, 11:52, 18:22

Impact

• Gateway memory grows from ~300MB (fresh start) to ~1.2GB over 5 days
• Requires weekly restart to prevent OOM

Expected Behavior

• Correct year calculation (2026, not 2025)
• Correct path resolution for CORE.md (workspace root, not memory-system/)
• Directory listing instead of file read for agent-tasks/

Additional Context

• These errors appear to be generated by OpenClaw internal cron/heartbeat mechanism, not user scripts
• The 8-hour interval suggests a scheduled task is triggering these reads