[CASSANDRA-10618] Read ghost data

• Type: Bug
• Status: Resolved
• Priority: Normal
• Resolution: Cannot Reproduce

Environment

Cassandra 2.1.2

• Severity: Normal
• Assignee: Unassigned
• Reporter: Zhao Yang
• Created: 29/Oct/15 08:14
• Updated: 16/Apr/19 09:30
• Resolved: 19/Nov/15 09:48

Description

In a table( pk, ck, value) with 1 partition key and 1 clustering key.

SELECT * FROM table where pk='PK1' AND ck1='CK1' (full primary keys) -> will return a row that doesn't appear in range scan query (SELECT * FROM table where pk='PK1').

I tested it with Java Driver 2.1.5 as well as DevCenter. Our environment has only 1 node and replication factor 1. It's our development DB, multiple developers are accessing it. And we are using client-side timestamp generator.

If I use nodetool to compact this table, this ghost data will disappear.

What can be the cause? because of un-order timestamp?

Thank you very much.

Issue Links

• duplicates: CASSANDRA-8558 deleted row still can be selected out