Fault Tolerance Strategy

Dubbo offers a variety of fault-tolerant scenarios when a cluster call fails, with a default failover retry.

The relationship between nodes:

• This Invoker is the callable Service's abstract of theProvider, and the Invoker packaging theProvider's address and Service's interface.
• The Directory represent multiple Invoker,You can think of it as List<Invoker>,But unlike List,its value can be dynamically changing.such as registry push changes
• The Cluster disguises multipleInvoker in Directory as aInvoker,The upper transparent, masquerade process contains fault-tolerant logic, call failed, try another
• The Router is responsible for selecting subsets according to routing rules from multiple Invokers, such as read-write separation, application isolation, etc.
• LoadBalance is responsible for selecting a specific one from multipleInvoker for this call. The selection process includes the load balancing algorithm. If the call fails, it needs to be re-selected

Cluster fault-tolerant mode

You can also customize the cluster fault tolerance strategy, see Cluster extension for more details.

Failover Cluster

Failure automatically switch, when there is failure, retry the other server (default). Usually used for read operations, but retries can result in longer delays. The times of retries can be set via retries =2 (excluding the first time).

The times of retries is configured as follows:

<dubbo:service retries="2" />

OR

<dubbo:reference retries="2" />

OR

<dubbo:reference>
    <dubbo:method name="findFoo" retries="2" />
</dubbo:reference>

Failfast Cluster

Fast failure, only made a call, failure immediately error. Usually used for non-idempotent write operations, such as adding records

Failsafe Cluster

Failure of security, anomalies, directly ignored. Usually used to write audit logs and other operations.

Failback Cluster

Failure automatically restored, failed to record the background request, regular retransmission. Usually used for message notification operations.

Forking Cluster

Multiple servers are invoked in parallel, returning as soon as one succeeds. Usually used for real-time demanding read operations, but need to waste more service resources. The maximum number of parallelism can be set with forks=2.

Broadcast Cluster

Calling all providers broadcast, one by one call, any error is reported (2.1.0+). It is usually used to notify all providers to update local resource information such as caches or logs.

Cluster mode configuration

Follow the example below to configure cluster mode on service providers and consumers

<dubbo:service cluster="failsafe" />

OR

<dubbo:reference cluster="failsafe" />