[CASSANDRA-7485] Missing data from secondary index when secondary index is against a compound PK member

• Type: Bug
• Status: Resolved
• Priority: Normal
• Resolution: Duplicate
• Fix Version/s: 2.0.13
• Component/s: Feature/2i Index
• Severity: Normal
• Assignee: Sam Tunnicliffe
• Reporter: Vahram Sukyas

Authors

Sam Tunnicliffe

• Created: 01/Jul/14 16:52
• Updated: 16/Apr/19 09:31
• Resolved: 22/Jan/15 17:43

Description

Under 2.0.8, assuming the following schema:

CREATE TABLE foo (
a text,
b text,
c text,
d text,
PRIMARY KEY (a, b)
);

CREATE INDEX foo_b_idx ON foo (b);

If you do a delete against c or d but specify a and b in the where clause, it will remove the secondary index reference to b (verified that it removes it from the underlying SSTable as well).

Example:

cqlsh:test> insert into foo (a, b, c, d) values ('test','test2','to_be_deleted', 'test3');
cqlsh:test> select * from foo;

a | b | c | d
-----------------------+------
test | test2 | to_be_deleted | test3

(1 rows)

cqlsh:test> delete c from foo where a = 'test' and b = 'test2';

cqlsh:test> select * from foo where b = 'test2';

(0 rows)

cqlsh:test> select * from foo where a = 'test' and b = 'test2';

a | b | c | d
--------------+------
test | test2 | null | test3

(1 rows)

cqlsh:test> select * from foo;

a | b | c | d
--------------+------
test | test2 | null | test3

(1 rows)

Dropping and recreating the index in CQL brings the secondary index reference back.

Issue Links

• duplicates: CASSANDRA-8206 Deleting columns breaks secondary index on clustering column