Problem

Memory index SQLite database (~/.openclaw/memory/main.sqlite) is created with journal_mode=delete (SQLite default). This causes frequent database corruption when the gateway receives SIGTERM during write operations (restart, update, config changes).

Impact

• Users see "database disk image is malformed" errors
• Memory search stops working until manual rm + openclaw memory index
• Happens repeatedly since gateway restarts are common

Root Cause

journal_mode=delete removes the journal file after each transaction. If the process is killed between deleting the journal and completing the write, the database corrupts.

Suggested Fix

Set PRAGMA journal_mode=WAL; when creating or opening the memory database. WAL mode:

• Is crash-safe (survives SIGTERM/SIGKILL mid-write)
• Allows concurrent reads during writes
• Is the recommended mode for applications that restart frequently

// After opening the database connection:
db.pragma('journal_mode = WAL');

Workaround

sqlite3 ~/.openclaw/memory/main.sqlite "PRAGMA journal_mode=WAL;"

But this resets if the DB is recreated (e.g., after openclaw memory index on a corrupted DB).

Environment

• OpenClaw 2026.3.2
• macOS arm64
• SQLite via better-sqlite3