Skip to content
Router.Homes

No internet or intermittent drops: all devices, one client, one app

Define which “no internet” this is

Section titled “Define which “no internet” this is”
Symptom More likely layer First step
All devices cannot browse WAN/upstream/DNS/NAT test WAN/DNS from the router
Only one device is offline that client’s IP/gateway/DNS/Wi-Fi/VPN compare another client on the same LAN
Only one app fails app DNS/proxy/QUIC/IPv6 try another browser, disable proxy, test DNS
Intermittent link, DHCP lease, Wi-Fi, upstream record a timeline; one ping is not enough
Ping IP works, web fails DNS or proxy nslookup + alternate DNS

Five layers

Section titled “Five layers”

1. WAN / upstream

Section titled “1. WAN / upstream”
Terminal window
ubus call network.interface.wan status
ip -4 addr show eth1
ip route
ping -c 3 192.168.50.1
ping -c 3 1.1.1.1
  • No WAN address: check DHCP/PPPoE (see 007);
  • Gateway unreachable: cable/upstream/ONT;
  • Gateway reachable but public IP fails: upstream/NAT/firewall;
  • Partial loss: record time/link state before changing config.

2. DNS

Section titled “2. DNS”
Terminal window
nslookup openwrt.org 127.0.0.1
nslookup openwrt.org 192.168.50.1
cat /tmp/resolv.conf.d/resolv.conf.auto
logread | grep -i dnsmasq | tail -20

With dnsmasq stopped in this lab:

  • router ping 1.1.1.1 still worked;
  • nslookup openwrt.org 127.0.0.1 failed;
  • client ping 1.1.1.1 worked, DNS failed;
  • restarting dnsmasq recovered.

That is the classic “IP works, names fail” pattern.

3. LAN / DHCP / client address

Section titled “3. LAN / DHCP / client address”
Terminal window
ip -4 addr show br-lan
cat /tmp/dhcp.leases
ip neigh

On the client:

Terminal window
ip addr show
ip route
cat /etc/resolv.conf
ping 192.168.1.1

Wrong gateway, IP conflict, or an old subnet all look like “only this device is offline”.

4. Firewall / NAT / forwarding

Section titled “4. Firewall / NAT / forwarding”
Terminal window
uci show firewall
nft list ruleset 2>/dev/null | head -80
logread | grep -i firewall | tail -20

If WAN has an address and DNS is fine but LAN clients cannot reach the internet, check:

  • LAN/WAN zones;
  • WAN masquerade;
  • LAN→WAN forward;
  • custom rules blocking DNS/forwarding.

5. One client / one app

Section titled “5. One client / one app”

One client:

  • remove its default route: ip route del default; if the gateway still works but public IP fails, the issue is that client’s route;
  • restore: ip route add default via 192.168.1.1;
  • check Wi-Fi, VPN, proxy, private DNS, IPv6 preference.

One app:

  • try another browser / disable proxy;
  • compare with nslookup/curl -v;
  • app-level DoH/QUIC/proxy can bypass the router DNS;
  • do not “reinstall the app” instead of layering the diagnosis.

Recorded demo (015 lab)

Section titled “Recorded demo (015 lab)”
015 lab: DNS, WAN and one-client routeDownload .cast
Real OpenWrt 25.12.6 VM + netns client: stop DNS (IP works/DNS fails), ifdown WAN (gateway works/public fails), then remove the client default route (only that client fails). Code blocks are authoritative.

Recording intermittent drops

Section titled “Recording intermittent drops”

Do not write “sometimes drops”. Record:

  • time, duration, whether it self-recovers;
  • all devices or one;
  • wired vs Wi-Fi, 2.4/5 GHz;
  • whether the WAN address/lease changed;
  • link/DHCP/DNS/firewall timestamps in logread;
  • whether it happens during download/upload/roaming/sleep.

Without a timeline, “intermittent” is hard to fix or verify.

Recovery order

Section titled “Recovery order”
  1. Restore WAN/DNS service first; do not factory-reset;
  2. For one client, change that client, not the router;
  3. For one app, disable proxy/private DNS first;
  4. If rebooting, record the state before reboot;
  5. Factory reset is the last resort; back up first.

E2 lab run (2026-10-20)

Section titled “E2 lab run (2026-10-20)”
  • Environment: OpenWrt 25.12.6 x86 VM 200; WAN DHCP 192.168.50.191/24; LAN 192.168.1.1/24; netns client 192.168.1.175/24.
  • DNS stopped: router public ping worked; router/client DNS failed; dnsmasq restart recovered.
  • WAN ifdown: WAN address disappeared; client pinged the gateway but not 1.1.1.1; ifup wan recovered.
  • Client default route removed: gateway worked, public IP unreachable; restoring the route recovered.
  • Not tested: real Wi-Fi, physical link flaps, upstream jitter, app proxy/DoH/QUIC, IPv6 preference.
  • Cast: public/casts/015-triage.cast; recipe: tests/recipes/no-internet-triage-25.12.md.

Sources and review

Section titled “Sources and review”
  • This is E2 partial: three controlled failures were tested; intermittent/physical/app layers were not.
  • No independent reviewer; conclusions are limited to this VM/netns environment.
  • Do not blame the ISP from one ping or one failed page load.

Sources

Router.Homes

Independent OpenWrt notes. Every page states the firmware, device, package manager and way back.

Icons: Solar by 480 Design, CC BY 4.0