In cockroachlabs/support#2583, we observed a corrupt timeseries key. The debug pebble scan command will surface all the top-level live keys within a range, but it won't tell us about all the constituent internal MERGE keys that make up a timeseries key. It might be useful to view all the internal MERGE keys corresponding to the timeseries keys. We might be observing a single internal MERGE key with a corrupted tail of a user key, causing the internal key to be treated as a new, unique user key. In the referenced issue, relevant data is probably all deleted or compacted away, but a scan-internal tool might've helped otherwise.
In cockroachlabs/support#2583, we observed a corrupt timeseries key. The
debug pebble scan
command will surface all the top-level live keys within a range, but it won't tell us about all the constituent internal MERGE keys that make up a timeseries key. It might be useful to view all the internal MERGE keys corresponding to the timeseries keys. We might be observing a single internal MERGE key with a corrupted tail of a user key, causing the internal key to be treated as a new, unique user key. In the referenced issue, relevant data is probably all deleted or compacted away, but a scan-internal tool might've helped otherwise.Jira issue: CRDB-31512