Originally from the User Slack
@sandr8: Are range queries over clustering columns potentially scanning a lot of rows because of having to leap over tombstones or this issue only affects Cassandra and ScyllaDB has a way to avoid it?
@avi: It affects ScyllaDB too but we have some mitigation. Most important is the ability to return empty pages, so if you have a really long run of tombstones, the query doesn’t time out.
@sandr8: Thank you! I would love to understand better what you mean by the ability to return empty pages.
@avi: Some more info here, but I don’t think we have a user level explanation: https://github.com/scylladb/scylladb/commit/e9cbc9ee85c25deb5ba9ee67ffa6e3ca9f904660
GitHub: Merge ‘Add support for empty replica pages’ from Botond Dénes · scylladb/scylladb@e9cbc9e