You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When investigating corruption, it can be useful to try to understand the lineage of a particular sstable. This can be painstakingly reconstructed from the pebble logs by grepping for file numbers to find compactions that constructed the sstable, repeating on compaction inputs. This forms a tree with the particular sstable at the root, compactions forming interior nodes, and flushes/ingests forming the leaves.
We could add a command to the debug tool to perform this search. The tree can be as deep as the beginning of the store's history, so we'll want to be able to prune it. It might be useful to prune by depth, but also by start timestamp (eg, if you know the timestamp of a key, you can search for tree leaves around the key's timestamp and younger).
When investigating corruption, it can be useful to try to understand the lineage of a particular sstable. This can be painstakingly reconstructed from the pebble logs by grepping for file numbers to find compactions that constructed the sstable, repeating on compaction inputs. This forms a tree with the particular sstable at the root, compactions forming interior nodes, and flushes/ingests forming the leaves.
We could add a command to the debug tool to perform this search. The tree can be as deep as the beginning of the store's history, so we'll want to be able to prune it. It might be useful to prune by depth, but also by start timestamp (eg, if you know the timestamp of a key, you can search for tree leaves around the key's timestamp and younger).
Motivated by cockroachdb/cockroach#114421.
Jira issue: PEBBLE-81
The text was updated successfully, but these errors were encountered: