Reliability & Robustness Improvements

This release addresses 14 identified reliability gaps to make VPN-AP truly resilient under all conditions.

Critical Fixes

• Watchdog now recovers VPN failures — VPN drops are detected via tunnel health checks (ping verification) and auto-reconnected
• Fixed infinite escalation loop — Escalation capped at 3/day; per-service counters only reset on success
• Eliminated VPN recovery race condition — Recovery consolidated into watchdog only (removed from portal health thread)
• Atomic firewall transitions — All iptables scripts use iptables-restore for zero-gap rule loading
• Fixed DNS redirect race condition — Firewall set up before DNS redirect removal during transitions

New Capabilities

• WiFi auto-recovery — Watchdog reconnects to last known WiFi on disconnect
• Exponential backoff — Recovery attempts use progressive delays (60s → 16min) to prevent storms
• Timeout protection — All systemctl (30s) and iptables (10s) calls wrapped with timeouts
• Hotel login auto-detection — Portal page polls for internet and shows success automatically
• Friendly WiFi error messages — Cryptic nmcli errors translated to plain English
• Prominent captive portal banner — Large banner with direct link when hotel login needed

Hardening

• systemd restart limits — Services stop restarting after repeated failures
• Management access rule dedup — Delete-before-insert pattern prevents iptables rule accumulation
• Improved log rotation — 3 history files (~4MB) instead of 1
• DNS resolution verification — dnsmasq health check verifies actual DNS resolution
• 2-of-3 ping checks — Upstream and internet checks use multiple targets for reliability

Files Changed

scripts/watchdog.sh · scripts/captive-portal-server.py · scripts/iptables-vpn-mode.sh · scripts/iptables-internet-mode.sh · scripts/iptables-captive-mode.sh · systemd/vpn-ap-watchdog.service · systemd/captive-portal.service · systemd/vpn-ap.service

Full Changelog: https://github.com/RFingAdam/VPN-AP/blob/main/CHANGELOG.md#140---2026-02-21