Does sharding-jdbc support dynamically adding data sources? My question is: a tenant corresponds to a DB, and how to dynamically update the sharding data source when adding a tenant during project operation?

[craftsman4j]

[liaoxianrong]

by updating ContextManager.metaDataContexts can dynamically change the data sources.
first, define a ContextManagerLifecycleListener to get the ContextManager:

public class CustomContextManagerLifecycleListener implements ContextManagerLifecycleListener {
	public static final Map<String, ContextManager> CONTEXT_MANAGER_HOLDER = new ConcurrentHashMap<>();
	@Override
	public void onInitialized(String databaseName, ContextManager contextManager) {
		CONTEXT_MANAGER_HOLDER.put(databaseName, contextManager);
	}

	@Override
	public void onDestroyed(String databaseName, InstanceType instanceType) {
		CONTEXT_MANAGER_HOLDER.remove(databaseName);
	}
}

and then, update metaDataContexts :

ContextManager contextManage = CONTEXT_MANAGER_HOLDER.get("logic_db");
ShardingSphereDatabase database = contextManager.getDatabase("logic_db");
		database.getRuleMetaData().getRules().forEach(rule -> {
			if (rule instanceof ShardingRule) {
				ShardingRule shardingRule = ((ShardingRule) rule);
				shardingRule.getShardingTables().forEach((name, shardingTable) -> {
					List<DataNode> actualDataNodes = shardingTable.getActualDataNodes();
					actualDataNodes.add(new DataNode(actualDataNodes.get(0).getDataSourceName(), "order_new"));
					actualDataNodes = new HashSet<>(actualDataNodes).stream().collect(Collectors.toList());
					Set<String> actualTables = (Set<String>) ReflectUtil.getFieldValue(shardingTable, "actualTables");
					actualTables.add("order_new");
					Map<DataNode, Integer> dataNodeIndexMap = new HashMap<>();
					Map<String, Collection<String>> dataSourceToTablesMap = shardingTable.getDataSourceToTablesMap();
					for (int i = 0; i < actualDataNodes.size(); i++) {
						dataNodeIndexMap.put(actualDataNodes.get(i), i);
						dataSourceToTablesMap.get(actualDataNodes.get(i).getDataSourceName()).add(actualDataNodes.get(i).getTableName());
					}
					ReflectUtil.setFieldValue(shardingTable, "dataNodeIndexMap", dataNodeIndexMap);
					ReflectUtil.setFieldValue(shardingTable, "dataSourceToTablesMap", dataSourceToTablesMap);
					ReflectUtil.setFieldValue(shardingTable, "actualTables", actualTables);
					ReflectUtil.setFieldValue(shardingTable, "actualDataNodes", actualDataNodes);

				});
				log.info("shardingRule:{}", shardingRule);
			}
		});

[yflytom]

比如我这个新增的数据源

，如何交给shardingjdbc

[liaoxianrong]

DataSource存放在metaDataContexts的ResourceMetaData，可以仿照上面的方法替换里面的内容：

@RequiredArgsConstructor
@Getter
public final class ResourceMetaData {
    
    private final Map<StorageNode, DataSource> dataSources;
    
    private final Map<String, StorageUnit> storageUnits;
    
    public ResourceMetaData(final Map<String, DataSource> dataSources) {
        this.dataSources = StorageNodeAggregator.aggregateDataSources(dataSources);
        Map<String, StorageNode> storageUnitNodeMap = dataSources.keySet().stream()
                .collect(Collectors.toMap(each -> each, StorageNode::new, (oldValue, currentValue) -> oldValue, LinkedHashMap::new));
        Map<String, DataSourcePoolProperties> dataSourcePoolPropsMap = dataSources.entrySet().stream().collect(
                Collectors.toMap(Entry::getKey, entry -> DataSourcePoolPropertiesCreator.create(entry.getValue()), (oldValue, currentValue) -> oldValue, LinkedHashMap::new));
        storageUnits = new LinkedHashMap<>();
        for (Entry<String, StorageNode> entry : storageUnitNodeMap.entrySet()) {
            storageUnits.put(entry.getKey(), new StorageUnit(entry.getValue(), dataSourcePoolPropsMap.get(entry.getKey()), dataSources.get(entry.getValue().getName())));
        }
    }

[yflytom]

好的，感谢老哥，我看看

[liaoxianrong]

其实一般来说，数据源是相对确定的，如果到了需要增加数据源的场景，那么大概率是数据库实例的扩容，比如增加一台单独服务器，这种情况下，重启服务重新创建datasource会更合适。而不是通过动态去创建datasource，毕竟datasource的维护是很谨慎的。

[yflytom]

我这里的场景是相当于多租户的场景，后续可能会存在，新增一个租户，由于隔离要求高，所以是单独的数据库

[linghengqian]

Does sharding-jdbc support dynamically adding data sources?

• There is no doubt that as long as you get the class instance of ContextManager, you can add, delete, modify, check and refresh the internal Metadata of ShardingSpehre, but through a complex Java API.
• I wrote a Chinese article 2 years ago, but there is no corresponding English article, and I don't have time to update the content of this article. Refer to [Article Contribution] Welcome to Leave Your Article Link Here #12258 (comment) .
• For how to get the class instance of ContextManager of a ShardingSphere JDBC DataSource in the life cycle of a microservice, refer to the unit test of ShardingSphere Agent.
• I would say that dynamically adjusting the configuration of ShardingSphere JDBC is a broad topic, but the core issue is that the internal APIs of DistSQL provided to ShardingSphere Proxy are changing significantly in each version, making it impossible to simply maintain the markdown documentation of these APIs.
• Considering that [Event] ShardingSphere is going to be a mentor community for GSoC 2024 #30252 (comment) was not selected in this year's GSoC 2024, I would say that assuming someone is willing to take over the relevant work to make DistSQL available on ShardingSphere JDBC?