backup-restore

name	backup-restore
description	Create backups before changes and restore if things go wrong. Triggers: "backup", "restore", "undo", "go back", "roll back", "save point".

Backup & Restore

MANDATORY: Always create a backup before any state-changing operation on the VPS.

Creating a Backup

Full System Backup

Backs up all workspace files, the config, and the credentials file. Think of it as a save point.

ssh -i ~/.ssh/id_ed25519_github root@72.62.213.231 \
  "tar -czf ~/.openclaw/backups/manual-$(date +%Y%m%d-%H%M%S).tar.gz \
   ~/.openclaw/workspace/ ~/.openclaw/openclaw.json ~/.openclaw/.env"

Tell Trent: "This creates a snapshot of everything. If something goes wrong, we can restore to exactly this state."

Single Agent Backup

If only one agent's files are changing:

ssh -i ~/.ssh/id_ed25519_github root@72.62.213.231 \
  "tar -czf ~/.openclaw/backups/<agent>-$(date +%Y%m%d-%H%M%S).tar.gz \
   ~/.openclaw/workspace/"

List Available Backups

ssh -i ~/.ssh/id_ed25519_github root@72.62.213.231 \
  "ls -lh ~/.openclaw/backups/ 2>/dev/null | sort -k6,7 || echo 'No backups directory found'"

Show Trent the list with dates and sizes. Explain: "Each of these is a save point you can restore from."

Restoring from Backup

Step 1: Choose the Backup

Show available backups and ask Trent which one to restore to.

Step 2: Stop the Gateway (YELLOW)

Tell Trent: "The agent will be offline for about 1-2 minutes during the restore."

ssh -i ~/.ssh/id_ed25519_github root@72.62.213.231 "systemctl stop openclaw-gateway"

Wait for confirmation before proceeding.

Step 3: Restore Files

ssh -i ~/.ssh/id_ed25519_github root@72.62.213.231 \
  "cd / && tar -xzf ~/.openclaw/backups/[BACKUP_FILE].tar.gz"

Step 4: Verify Permissions

After restore, always fix permissions (the tar might not preserve them correctly):

ssh -i ~/.ssh/id_ed25519_github root@72.62.213.231 \
  "chmod 600 ~/.openclaw/.env && chmod 444 ~/.openclaw/openclaw.json"

Step 5: Restart and Verify

ssh -i ~/.ssh/id_ed25519_github root@72.62.213.231 "systemctl start openclaw-gateway"

Wait 15 seconds, then dispatch the deploy-verifier agent to confirm everything is healthy.

Step 6: Confirm with Trent

"Restore complete. The system is back to the state from [backup date/time]. Everything looks healthy."

Or if verification failed: "The restore finished but there's an issue. [explain]. Would you like to try a different backup or should we troubleshoot?"

Checking Backup Health

Periodically check that backups exist and aren't too old:

ssh -i ~/.ssh/id_ed25519_github root@72.62.213.231 \
  "ls -lt ~/.openclaw/backups/ 2>/dev/null | head -5"

If the most recent backup is more than 7 days old, suggest creating a fresh one.

Rules

• NEVER restore without stopping the gateway first.
• NEVER delete the only remaining backup.
• After restore, ALWAYS verify permissions on .env (600) and openclaw.json (444).
• After restore, ALWAYS run a health check to confirm recovery.
• Explain every step to Trent before executing.
• If the backups directory doesn't exist, create it: mkdir -p ~/.openclaw/backups/