openclaw-gateway-hook

Systemd service (openclaw-gateway-hook.service) running /opt/scripts/openclaw-gateway-hook.sh. Watches Docker start events on repo-openclaw-gateway-1 and runs /opt/scripts/openclaw-token-sync.sh 10s after each gateway start. Logs to /var/log/openclaw-watchdog.log with tag [gateway-hook].

2026-07-26 — dead agent-restart step removed

The hook previously also ran docker restart openclaw-a72o-openclaw-1 after token sync. That agent container was retired with the v3 gateway-only architecture (Kulu removed 2026-04-13), so every gateway start since then logged No such container + ERROR: Failed to restart agent container (seen 07-22, 07-24, 07-26). The restart step was removed (token sync kept, now logs “Token sync complete”), the systemd unit description was updated to match, and the service was restarted. Committed in /opt/scripts as 68d74ef; the 02:45 push cron will sync it to origin.

2026-07-27 — wa_wrong_account incident resolution (related)

The 7/26 evening Kilroy WhatsApp pairing retry mis-paired Dustin’s personal account (+12028683616) onto Rob’s gateway, overwriting Rob’s creds (wrong creds quarantined same evening at config/credentials/whatsapp.WRONG-ACCOUNT-QUARANTINE-20260726T194210). On 7/27, Rob’s +16672742864 session was restored from personal-config-2026-07-26_0200.tar.gz — the provider started as the correct account but WhatsApp’s server rejected the session (“session logged out”): the device slot was invalidated, so no headless recovery exists. Result: the wa_wrong_account CRIT loop auto-resolved (watchdog greps the last “starting provider” line, now correct), and a single correct wa_channel_down alert went to Dustin. Rob’s WhatsApp stays down until Dustin re-links with openclaw channels login --channel whatsapp using the phone that owns +16672742864.