routerd soak monitor - 2026-05-13

Scope

This note tracks the Phase 3.Z soak across:

Host	OS target	Role
homert02	Ubuntu	production home router
router02	NixOS	lab router
router04	FreeBSD	lab router

The soak samples every five minutes into /var/log/routerd-soak-YYYYMMDD.log using scripts/routerd-soak-sample.sh.

Signals

Each sample captures:

• routerctl status -o json
• routerctl resources -o json
• routerctl wireguard show wg-mesh -o json
• routerctl tailscale peers -o json
• routerd process RSS and elapsed time when ps supports it
• UDP socket snapshot when ss is present

Initial state

The monitor is intended to detect generation drift, resource-count churn, WireGuard handshake staleness, Tailscale peer-state regressions, and obvious process RSS growth. The 24h result is pending until the next collection window has elapsed.

Initial deployment on 2026-05-13 UTC:

Host	Scheduler	Initial result
homert02	root cron, every 5 minutes	routerctl status Healthy, resource count 94, wg-mesh handshakes fresh, Tailscale peer list available
router02	runtime systemd timer, every 5 minutes	routerctl status Healthy, wg-mesh handshakes fresh, Tailscale skipped because tailscale is not installed
router04	root cron, every 5 minutes	routerctl status Healthy, wg-mesh handshakes fresh, Tailscale skipped because tailscaled is not running

NixOS router02 uses a runtime systemd timer under /run/systemd/system because /etc/systemd/system is read-only on that host. homert02 and router04 use root crontab entries.

Follow-up checklist

• Confirm each host appends samples for at least 24h.
• Compare first and last resource counts.
• Check that Healthy/Applied resources do not flap.
• Check wg-mesh peer handshakes remain fresh on Linux and FreeBSD peers.
• Record any anomaly here before making a larger follow-up task.