← Back to Operations and recoveryOpenClaw operations
Logs, status, and health checks
The point of status and logs is not to sound informed. It is to know what is actually true right now.
Health ruleWeak status text, active labels, and silence are not enough by themselves. Use evidence that corresponds to the claimed state.
Weak vs strong signals
What should and should not count as evidence?
Weak signal: repeated status text“Still working” means almost nothing without a new artifact or log event.
Weak signal: active labelsAn “active” queue row is not proof if no downstream change landed.
Weak signal: silenceSilence only means failure when a real checkpoint should already exist.
Strong signal: direct artifact or outputFiles, URLs, logs, screenshots, and process state are much stronger.
Useful examples
Examples of honest operational checks
Live route checkUse HTTP status and returned content, not just “deploy said complete.”
Queue health checkLook for real next artifacts, not only labels in a schedule or queue file.
Runtime health checkUse logs and process state when the runtime itself might be unhealthy.
Open deeper reference links for this page