A replica is a copy of database data stored on another node. Multiple replicas improve availability and fault tolerance while helping distribute read and query traffic. section 1.1.2 section 1.1.3
ScyllaDB stores each row in multiple locations according to the configured replication factor. Every node is equal, and requests are forwarded to the nodes that own the data. section 1.2.1
The required number of replicas affects availability and consistency:
- ALL requires every replica to acknowledge the query, providing the strongest requirement but tolerating no node failure. section 1.2.1
- QUORUM requires a majority of replicas, balancing consistency with tolerance for a failed node. section 1.2.1
- ONE requires a single replica, improving availability while increasing the chance of inconsistent results. section 1.2.1
Three replicas
With three replicas, ALL needs three successful responses and tolerates zero failures; QUORUM needs two and tolerates one; ONE needs one and tolerates two. section 1.2.1
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 replica acknowledgement requirements, successful responses, and tolerated failures for ALL, QUORUM, and ONE. section 1.2.1
