QUORUM is a consistency level that requires a majority of replicas holding a key to acknowledge a read or write. In a three-replica cluster, two replicas must respond. This sits between ONE, which favors availability, and ALL, which requires every replica and cannot tolerate a node loss. section 1.2.1 section 1.4
Table 1.1 Sample of consistency level options, assuming a cluster with three replicas
|
Consistency level
|
Description
|
Number required to succeed
|
Failures tolerated
|
|---|---|---|---|
ALL
|
Requires all replicas to succeed
|
3
|
0
|
QUORUM
|
Requires a majority of replicas to succeed
|
2
|
1
|
ONE
|
Requires a single replica to succeed
|
1
|
2
|
Compare the acknowledgments and node failures tolerated by ALL, QUORUM, and ONE in a three-replica cluster. section 1.2.1
A coordinator receives the client request, contacts the replicas for the key, gathers their results, and checks whether enough replies satisfy the selected consistency level. Using QUORUM for both writes and reads means a majority receives each update and a read consults a majority, reducing the chance of inconsistent results. section 1.2.1
Reading reviews with one node down
With three nodes and one unavailable, a query for Ernie’s Eats can still succeed at QUORUM because the remaining two nodes form a majority. The same query at ALL fails because every node must respond. QUORUM is stronger than ONE: a lone reachable replica during a network partition could hold divergent data. section 2.4.2
Choose QUORUM when its trade-off fits the request:
- Use it to tolerate the loss of one node in a three-replica setup while still requiring two acknowledgments. section 1.2.1
- Use the same level for reads and writes when majority visibility of updates is needed. section 1.2.1
- Consider ONE when preserving service through more failures matters more than stronger consistency; consider ALL when every replica must respond. section 1.2.1 section 1.2.6
