[Bug] Dashboard unreachable + gateway instability on Windows in 2026.2.22-2 (fixed by downgrading to 2026.2.21-2)
Summary
On Windows 11, upgrading to openclaw 2026.2.22-2 caused dashboard/web UI instability and repeated gateway failures.
Symptoms:
- Dashboard/web UI intermittently unreachable from both localhost and tailnet
http://127.0.0.1:18789/ and https://<tailnet-host>:8443/ often returned 404 Not Found
openclaw gateway status frequently reported service stopped / exited immediately
- CLI failures with websocket close error:
gateway closed (1006 abnormal closure (no close frame))
- In failing windows,
curl to 127.0.0.1:18789 returned connect failures despite intermittent listeners
Downgrading to 2026.2.21-2 restored stable behavior immediately.
Environment
- OS: Windows 11 (Windows_NT 10.0.26200 x64)
- Node: 22.15.0
- OpenClaw bad version:
2026.2.22-2
- OpenClaw stable version:
2026.2.21-2
- Gateway mode: local loopback (
127.0.0.1:18789)
- Tailnet exposure:
tailscale serve mapping 8443 -> 18789
Repro (intermittent but frequent)
- Install/upgrade to
openclaw@2026.2.22-2 on Windows
- Start gateway service via
openclaw gateway start
- Open dashboard URL (tokenized) or tailnet route on
:8443
- Observe one or more of:
- dashboard returns 404
- UI fails to connect
openclaw gateway status flips to stopped
- CLI shows ws close
1006 abnormal closure
Expected behavior
- Gateway service remains running
- Dashboard route is consistently available
- No repeated
1006 websocket closure from local loopback gateway
Actual behavior
- Gateway appears to crash/stop intermittently
- Dashboard route becomes unavailable/404 during failures
- CLI/Gateway sessions show websocket abnormal closure
Evidence snippets
From debug output captured during failure:
Failed to start CLI: Error: gateway closed (1006 abnormal closure (no close frame))
Runtime: stopped (state Ready, last run time ...)
RPC probe: failed
curl: (7) Failed to connect to 127.0.0.1 port 18789 ... Could not connect to server
Workaround
Downgrade and pin:
openclaw gateway stop
npm uninstall -g openclaw
npm i -g openclaw@2026.2.21-2
openclaw gateway start
After downgrade:
- dashboard restored
- gateway status stable/running
Impact
High: Blocks dashboard/web UI and makes gateway reliability unpredictable on affected Windows setup.
Request
Please investigate regression between 2026.2.21-2 and 2026.2.22-2 around:
- gateway service lifecycle on Windows (scheduled task / process ownership)
- dashboard route availability under local loopback
- websocket connection handling causing
1006 abnormal closure
[Bug] Dashboard unreachable + gateway instability on Windows in 2026.2.22-2 (fixed by downgrading to 2026.2.21-2)
Summary
On Windows 11, upgrading to
openclaw 2026.2.22-2caused dashboard/web UI instability and repeated gateway failures.Symptoms:
http://127.0.0.1:18789/andhttps://<tailnet-host>:8443/often returned404 Not Foundopenclaw gateway statusfrequently reported servicestopped/ exited immediatelygateway closed (1006 abnormal closure (no close frame))curlto127.0.0.1:18789returned connect failures despite intermittent listenersDowngrading to
2026.2.21-2restored stable behavior immediately.Environment
2026.2.22-22026.2.21-2127.0.0.1:18789)tailscale servemapping8443 -> 18789Repro (intermittent but frequent)
openclaw@2026.2.22-2on Windowsopenclaw gateway start:8443openclaw gateway statusflips to stopped1006 abnormal closureExpected behavior
1006websocket closure from local loopback gatewayActual behavior
Evidence snippets
From debug output captured during failure:
Failed to start CLI: Error: gateway closed (1006 abnormal closure (no close frame))Runtime: stopped (state Ready, last run time ...)RPC probe: failedcurl: (7) Failed to connect to 127.0.0.1 port 18789 ... Could not connect to serverWorkaround
Downgrade and pin:
After downgrade:
Impact
High: Blocks dashboard/web UI and makes gateway reliability unpredictable on affected Windows setup.
Request
Please investigate regression between
2026.2.21-2and2026.2.22-2around:1006 abnormal closure