ShardingSphere built-in provides a variety of load balancer algorithms, including polling algorithm, random access algorithm and weight access algorithm, which can meet users’ needs in most business scenarios.
Moreover, considering the complexity of the business scenario, the built-in algorithm also provides an extension mode. Users can implement the load balancer algorithm they need based on SPI interface.
| Type | Describe | Limitations |
|---|---|---|
| ROUND_ROBIN | Within the transaction, read query are routed to the primary, and outside the transaction, the round-robin strategy is used to route to the replica | |
| RANDOM | Within the transaction, read query are routed to the primary, and outside the transaction, the random strategy is used to route to the replica | |
| WEIGHT | Within the transaction, read query are routed to the primary, and outside the transaction, the weight strategy is used to route to the replica | Attributes need to be configured, attribute name: ${replica-name}, data type: double, attribute name uses the name of the replica, and the parameter fills in the weight value corresponding to the replica. Weight parameter range min > 0, total <= Double.MAX_VALUE. |
| TRANSACTION_RANDOM | Display/non-display open transaction, read query are routed to multiple replicas using random strategy | |
| TRANSACTION_ROUND_ROBIN | Display/non-display open transaction, read query are routed to multiple replicas using round-robin strategy | |
| TRANSACTION_WEIGHT | Display/non-display open transaction, read query are routed to multiple replicas using weight strategy | Attributes need to be configured, attribute name: ${replica-name}, data type: double, attribute name uses the name of the replica, and the parameter fills in the weight value corresponding to the replica. Weight parameter range min > 0, total <= Double.MAX_VALUE. |
| FIXED_REPLICA_RANDOM | Open transaction displayed, and the read query is routed to a fixed replica using random strategy; otherwise, each read traffic is routed to a different replica using random strategy | |
| FIXED_REPLICA_ROUND_ROBIN | Open transaction displayed, and the read query is routed to a fixed replica using round-robin strategy; otherwise, each read traffic is routed to a different replica using round-robin strategy | |
| FIXED_REPLICA_WEIGHT | Open transaction displayed, and the read query is routed to a fixed replica using weight strategy; otherwise, each read traffic is routed to a different replica using weight strategy | Attributes need to be configured, attribute name: ${replica-name}, data type: double, attribute name uses the name of the replica, and the parameter fills in the weight value corresponding to the replica. Weight parameter range min > 0, total <= Double.MAX_VALUE. |
| FIXED_PRIMARY | All read query are routed to the primary |
rules:
- !READWRITE_SPLITTING
dataSources:
readwrite_ds:
staticStrategy:
writeDataSourceName: write_ds
readDataSourceNames:
- read_ds_0
- read_ds_1
loadBalancerName: random
loadBalancers:
random:
type: RANDOM