Real tailnet IPs and Telegram bot handles were being committed in docs/
memories/skills. Scrubbed all tracked markdown to ${VAR} placeholders; real
values now live in fleet.env (gitignored) and stay live via 'tailscale status'.
- add fleet.env.example (committed) + fleet.env (gitignored); .gitignore *.env
- AGENTS.md + HOST-MATRIX: masking convention so it can't recur
- also: domedog registered as Colibri agent (image-render/ffmpeg/build lane);
correct CAPABILITY-ROUTING example to real registered caps (domedog headless)
Past commits not rewritten (history moves to Codeberg at v1.0); this fixes HEAD.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
3.4 KiB
3.4 KiB
Network / SSH / tmux lag triage reference
Use this when a session investigates interactive lag, Wi-Fi quality, suspected RF/interference from a projector/peripheral, or whether kernel Wi-Fi drivers/firmware are at fault.
Evidence hierarchy
- Confirm the Wi-Fi stack:
lspci -nnk | sed -n '/Network controller/,+8p'nmcli -f GENERAL,WIFI-PROPERTIES,IP4 device show <iface>modinfo <driver>journalctl -k --since today --no-pager | egrep -i 'iwlwifi|wlp|wlan|firmware|deauth|disconnect|timeout|fail|error'
- Check package freshness only against configured repos, not against vague "newest upstream":
apt-cache policy firmware-iwlwifi linux-image-amd64 wireless-regdb network-manager iwapt list --upgradable 2>/dev/null | egrep 'linux-image|firmware|wireless|network-manager|iw'
- Separate harmless firewall/discovery noise from the interactive path:
UFW BLOCKfrom router DNS or LAN discovery ports 1900/5353/5355/3702 can be normal background noise.- Do not treat it as root cause unless it correlates with retransmits, loss, or the exact failing flow.
- Inspect actual SSH sockets:
ss -nti '( sport = :22 or dport = :22 )'- Strong fields:
rtt:<avg>/<var>,bytes_retrans,retrans:<active>/<total>,dsack_dups,reord_seen,rcv_ooopack,cwnd.
- Compare network layers:
- router/default gateway ping for local Wi-Fi/AP jitter
- public IP ping for ISP/internet jitter
- overlay/VPN peer ping if Tailscale/WireGuard is involved
- Inspect Wi-Fi conditions:
nmcli -f ACTIVE,SSID,BSSID,CHAN,RATE,SIGNAL,BARS,SECURITY dev wifi list --rescan yesiw dev <iface> linkip -s link show <iface>- 2.4 GHz channels plus weak/mid signal often explain sticky SSH even with zero packet loss.
- For suspected projector/peripheral interference, run a before/after capture with identical duration and compare:
- gateway RTT max/mdev
- SSH retransmits/reordering before vs after
- Wi-Fi signal/rate/channel before vs after
- packet capture only if tshark/tcpdump is installed and useful
Reusable scripts
This skill includes:
scripts/network-lag-baseline.sh: single snapshot collector for Wi-Fi, SSH sockets, pings, Tailscale state, and recent logs.scripts/network-interference-capture.sh: timed before/after test with ping streams and optional tshark/tcpdump packet capture.
Copy scripts out or run them from the skill directory. Prefer setting env vars rather than hardcoding hosts:
WIFI_IFACE=wlp1s0 ROUTER_IP=192.168.1.1 REMOTE_IP=${DOMEDOG_TS_IP} ./network-lag-baseline.sh ./baseline.txt
WIFI_IFACE=wlp1s0 ROUTER_IP=192.168.1.1 REMOTE_IP=${DOMEDOG_TS_IP} ./network-interference-capture.sh 90 ./before-projector
WIFI_IFACE=wlp1s0 ROUTER_IP=192.168.1.1 REMOTE_IP=${DOMEDOG_TS_IP} ./network-interference-capture.sh 90 ./after-projector
Reporting pattern
Report in this order:
- Driver/firmware status and whether configured repositories offer newer packages.
- Kernel/journal findings: explicitly distinguish driver crashes/deauths from unrelated boot warnings or firewall noise.
- Wi-Fi signal/band/channel and local gateway jitter.
- SSH socket evidence: retransmits/reordering/RTT variance.
- Overlay path comparison, if present.
- Next experiment: before/after test or packet capture.
Avoid jumping to fixes like changing drivers, disabling firewalls, or changing power settings until the above evidence points there.