[CASSANDRA-16835] Scrub still uses "row" to mean "partitions", and has broken code

• Type: Bug
• Status: Resolved
• Priority: Normal
• Resolution: Fixed
• Fix Version/s: 3.11.12, 4.0.1, 4.1-alpha1, 4.1
• Component/s: Tool/nodetool
• Bug Category: Correctness
• Severity: Low
• Complexity: Low Hanging Fruit
• Discovered By: User Report
• Platform: All

Source Control Link

https://github.com/apache/cassandra/commit/e581a85b93acff0fecef7d41d9a94a2f89f810ba

Test and Documentation Plan

https://issues.apache.org/jira/browse/CASSANDRA-16835?focusedCommentId=17396932&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-17396932

• Assignee: Ekaterina Dimitrova
• Reporter: Ekaterina Dimitrova

Authors

Ekaterina Dimitrova, Sylvain Lebresne

Reviewers

Brandon Williams

• Created: 09/Aug/21 20:45
• Updated: 27/May/22 19:25
• Resolved: 17/Aug/21 19:23

Description

2 issues in the scrub code:
1) It still uses "row" to mean "partition". And not only in the code, but also in user messages. As we've fairly systematically remove such instances elsewhere in 3.0+, having it in scrub is going to confuse users which will almost surely misinterpret the results. If scrub says that it dropped 2 unreadable "rows" from your sstable, you might be ok with that when we're actually talking about CQL rows, but not if we talk of 2 full partitions.
2) There is a branch at the end of scrub that is supposed to handle the case where scrubbing a sstable generates no output at all (the sstable is completely hosed usually), mostly providing a more user friendly message. The code is broken (and has been for a long time, since CASSANDRA-7066 I believe) however such that this branch can simply never be taken (even when it should). While admittedly pretty minor, no reason to leave it that way.