Is it possible to merge read-write rule and discovery rule

[zhaojinchao95]

Hi community,

When i test use High availability MGR. It doesn't work well, i execute read or write operation，it always choose primary datasources. I see source code found it

@RequiredArgsConstructor
public final class DatabaseDiscoveryDataSourceRouter {
    
    private final DatabaseDiscoveryDataSourceRule rule;
    
    /**
     * Route.
     * 
     * @return data source name
     */
    public String route() {
        return rule.getPrimaryDataSourceName();
    }
}

Then, I try to use High availability MGR with ReadWriteRules. It's work well, my configuration is:

schemaName: database_discovery_db

dataSources:
  ds_0:
    url: jdbc:mysql://127.0.0.1:13308/demo_primary_ds?serverTimezone=UTC&useSSL=false
    username: root
    password: 
    connectionTimeoutMilliseconds: 3000
    idleTimeoutMilliseconds: 60000
    maxLifetimeMilliseconds: 1800000
    maxPoolSize: 50
    minPoolSize: 1
  ds_1:
    url: jdbc:mysql://127.0.0.1:13309/demo_primary_ds?serverTimezone=UTC&useSSL=false
    username: root
    password: 
    connectionTimeoutMilliseconds: 3000
    idleTimeoutMilliseconds: 60000
    maxLifetimeMilliseconds: 1800000
    maxPoolSize: 50
    minPoolSize: 1
  ds_2:
    url: jdbc:mysql://127.0.0.1:13308/demo_primary_ds?serverTimezone=UTC&useSSL=false
    username: root
    password: 
    connectionTimeoutMilliseconds: 3000
    idleTimeoutMilliseconds: 60000
    maxLifetimeMilliseconds: 1800000
    maxPoolSize: 50
    minPoolSize: 1

rules:
- !READWRITE_SPLITTING
  dataSources:
    rd_rs:
      autoAwareDataSourceName: pr_ds

- !DB_DISCOVERY
  dataSources:
    pr_ds:
      dataSourceNames:
        - ds_0
        - ds_1
        - ds_2
      discoveryTypeName: mgr
  discoveryTypes:
    mgr:
      type: MGR
      props:
        groupName: b13df29e-90b6-11e8-8d1b-525400fc3996
        zkServerLists: 'localhost:2181'
        keepAliveCron: '0/50 * * * * ?'

Through the MGR we know which one are primary datasource (writeDataSourceName) and other slave datasources (readDataSourceNames), So, i think - !READWRITE_SPLITTING and - !DB_DISCOVERY should be merged, and let DB_DISCOVERY be an augmented version of READWRITE

[zhaojinchao95]

@terrymanu @tristaZero @menghaoranss @strongduanmu Hi, we can discuss it together!

[strongduanmu]

@ZjcNB I like this suggestion because DatabaseDiscovery looks different from other DistributedRule, such as ReadWriteSplitting and Shadow. It does not have routing capabilities. It looks more like an implementation of ReadWriteSplitting now, which is different from the traditional master-slave architecture.

[zhaojinchao95]

After discussion, maybe some users will use the discovery module for strong data consistency, so let's continue to keep it.