Cassandra Read inconsistency

1
votes

I m using Cassandra as Production Database for my company , i m having a read Consistency Problem. The Records per day is Supposed to be 1440 , on a first read it displays on 600-700 odd records on second read it displays 1440+ records.I m having 3 nodes in the cluster, and 3 secondary indexes, please help me out. I have used the Below cassandra version and configuration for write and Read.

Cassandra Version:
1.1.4 

Thrift Version:
070

Consistency level 
for Read  - ALL
for Write - ANY

I have done couple of try's including Increasing the timeout time , also have Checked running Nodetool repair in each nodes still haven't got any change in the consistency.

2
cassandra
Read means? multiget or range_slice or what type of read? – Tamil

2 Answers

6
votes

Le douard is close: W=QUORUM + R=QUORUM would give you strong consistency. So would W=ONE + R=ALL. But not W=ANY, which is a "special" level allowing the cluster to accept writes even when all the "natural" replicas (that are consulted by reads) are down.

See http://www.datastax.com/docs/1.1/dml/data_consistency.

4
votes

You are getting your consistency wrong. To have consistency in cassandra you need to read and write at least to Concistency level QUORUM