Database Reference
In-Depth Information
to which {C, D, E} are connected fails, then you have a network partition that isolates the
two subclusters {A, B} and {C, D, E}.
Cassandra is a fault-tolerant database, and network partitions are one such fault. As such, it
is able to continue operating in the face of a network partition and merge data in replication
once the partition is healed again.
Parttiittiioner
The partitioner controls how your data is distributed over your nodes. In order to find a
set of keys, Cassandra must know what nodes have the range of values you're looking for.
There are three types of partitioner: random partitioner, which is the default; order-pre-
serving partitioner; and collating order-preserving partitioner. You configure this in storage-
conf.xmlor cassandra.yaml(for 0.7) using the
<Partitioner>
element:
<Partition-
er>org.apache.cassandra.dht.RandomPartitioner</Partitioner>
. Note that parti-
tioning applies to the sorting of row keys, not columns.
Once you have chosen a partitioner type, you cannot change it without destroying your data
(because an
SSTable
is immutable).
See also
Order-Preserving Partitioner
and
Random Partitioner
.
Quorum
A majority of nodes that respond to an operation. This is a configurable consistency level.
In a quorum read, the proxy waits for a majority of nodes to respond with the same value.
This makes for a slower read operation, but also helps ensure that you don't get returned stale
data.
Rack--Aware S
e Sttrattegy
Random P
m Parttiittiioner
where to place the keys on the node ring. This has the advantage of spreading your keys
evenly across your cluster, but the disadvantage of causing inefficient range queries. This is
the default partitioner.
See also
Partitioner
and
Order-Preserving Partitioner
.
Range S
e Slliice
Query to get a subset of columns for a range of keys.
