[CASSANDRA-6004] Performing a "Select count(*)" when replication factor < node count causes assertion error and timeout

• Type: Bug
• Status: Resolved
• Priority: Normal
• Resolution: Fixed
• Fix Version/s: 2.0.1
• Component/s: Legacy/CQL

Environment

Two node setup
Ubuntu Server 12.04
Tested on JDK 1.6 and 1.7

• Severity: Normal
• Since Version: 2.0.0
• Assignee: Sylvain Lebresne
• Reporter: James P

Authors

Sylvain Lebresne

Reviewers

Jonathan Ellis

• Created: 11/Sep/13 02:04
• Updated: 16/Apr/19 09:32
• Resolved: 16/Sep/13 15:24

Description

When performing a "Select Count()" query on a table belonging to a keyspace with a replication factor less than the total node count, the following error is encountered which ultimately results in an rpc_timeout for the request:

ERROR 18:47:54,660 Exception in thread Thread[Thread-5,5,main]
java.lang.AssertionError
at org.apache.cassandra.db.filter.IDiskAtomFilter$Serializer.deserialize(IDiskAtomFilter.java:116)
at org.apache.cassandra.db.RangeSliceCommandSerializer.deserialize(RangeSliceCommand.java:247)
at org.apache.cassandra.db.RangeSliceCommandSerializer.deserialize(RangeSliceCommand.java:156)
at org.apache.cassandra.net.MessageIn.read(MessageIn.java:99)
at org.apache.cassandra.net.IncomingTcpConnection.receiveMessage(IncomingTcpConnection.java:148)
at org.apache.cassandra.net.IncomingTcpConnection.handleModernVersion(IncomingTcpConnection.java:125)
at org.apache.cassandra.net.IncomingTcpConnection.run(IncomingTcpConnection.java:73)

The issue is not encountered when the replication factor is >= node count

To replicate the issue:
1) Create the keyspace: CREATE KEYSPACE demodb WITH REPLICATION =

{'class' : 'SimpleStrategy', 'replication_factor': 1}

;

2) Create the table CREATE TABLE users (
user_name varchar,
password varchar,
gender varchar,
session_token varchar,
state varchar,
birth_year bigint,
PRIMARY KEY (user_name));

3) Do a CQL query: "SELECT count( * ) FROM demodb.users" ;

The issue is reproducible even if the table is empty. Both CQLSH and client (astyanax) api calls are affected. Tested on two different clusters (2-node and 8-node)