Incident log
An incident is anything that went wrong in production, such as a wrong burn, a missed burn, a failed public audit or an exposed key. The policy says what counts and how each one is written up.
Recorded incidents
No activity yet
No incidents are recorded. The log is append-only. An entry is never edited after the fact, and a correction is a new entry that names the one it corrects.
Caught before it mattered
These problems were found before any coin depended on the part that failed. Nothing was signed, sent or burned because of them. They are listed so the record is complete.
A keeper kept restarting
- What happened
- After the switch to mainnet on Oct 11, a keeper on the host refused to start without the setting that allows mainnet sends, and the host's service manager restarted it again and again. It stopped before its first cycle each time, so it never signed or sent anything.
- What was done
- The cause was written up with the two ways out: stop the service, or start it with the mainnet setting when the run sheet calls for it. The keeper's own refusal is the safety check working as designed.
The status page could not find its data file
- What happened
- The status page loaded, but its data file returned a not-found error. The keeper that writes the file was not running, and a failed copy on the host could also have been hidden, because the publish step reported success even when a copy failed.
- What was done
- The publish step now fails when a copy fails and refuses to run as the root user. A check logs a line when the keeper's file goes missing or stale and another when it is back, and a probe script walks every link from the keeper to the page.