Hello! I have assisted in troubleshooting and found the issue: Your OpenClaw Gateway failed to start because the value of the OPENCLAW_GATEWAY_BIND environment variable was rejected by OpenClaw, causing the container to fall back to the fallback page. This is the reason for the 502 error.
Diagnosis Results
Key message from the pod logs:
=== Starting OpenClaw Gateway ===
2026-04-21T13:47:15.984+00:00 [gateway] loading configuration…
2026-04-21T13:47:25.196+00:00 Invalid --bind (use "loopback", "lan", "tailnet", "auto", or "custom")
=== Gateway exited (code: 0), restarting in 1h... ===
Fallback page serving on port 8080
Key point: The Gateway process reports Invalid --bind and exits, then runs a fallback static page on port 8080. From the K8s perspective, the Pod is 1/1 Running (because the fallback page responds), but the actual OpenClaw process has crashed. The Gateway is set to retry in 1 hour, which is why you see a persistent 502.
Your current setting:
OPENCLAW_GATEWAY_BIND=lan
In theory, lan is a valid value (one of the 5 valid values listed in the error message), but it is clearly being rejected in your current OpenClaw version/container environment. The most likely reason is that there is no explicit "lan" network interface definition inside the K8s container, so OpenClaw rejects this value.
Solution: Change to auto
Please change the environment variable to:
OPENCLAW_GATEWAY_BIND=auto
Steps:
- Go to Zeabur Dashboard → your
OpenClaw service → Variables tab
- Find the
OPENCLAW_GATEWAY_BIND entry
- Change the value from
lan to auto
- Save and click Restart in the top right corner to restart the service
auto means "OpenClaw determines the appropriate interface to bind to itself," which is the most suitable setting for container deployment environments (loopback binds to 127.0.0.1 and is only accessible from within the container, lan has an ambiguous definition in K8s environments, and tailnet requires Tailscale, while custom requires additional configuration).
Why not loopback?
Although loopback is also valid, it binds to 127.0.0.1, which is only accessible from within the pod — the Zeabur ingress cannot connect from the outside, so you would still get a 502. auto allows OpenClaw to bind to 0.0.0.0, enabling the ingress to route traffic correctly.
Expected Recovery Time
About 30–60 seconds after restarting the service, your https://skyky.zeabur.app/ should return to normal.
If you still get a 502 after changing to auto and restarting, please paste the new pod logs (searching for keywords like gateway, error, or listening), and we will continue troubleshooting.
