12. sharding-jdbc源碼之強制路由

阿飛Javaer，轉載請註明原創出處，謝謝！！

源碼分析

位於sharding-jdbc-core模塊下的包com.dangdang.ddframe.rdb.sharding.hint中，核心類HintManagerHolder的部分源碼如下：

/**
 * Hint manager holder.
 * <p>Use thread-local to manage hint.</p>
 * @author zhangliang
 */
@NoArgsConstructor(access = AccessLevel.PRIVATE)
public final class HintManagerHolder {

    // hint特性保存數據的核心變量，即保存一個HintManager類型對象到ThreadLocal中
    private static final ThreadLocal<HintManager> HINT_MANAGER_HOLDER = new ThreadLocal<>();

    /**
     * Set hint manager.
     * @param hintManager hint manager instance
     */
    public static void setHintManager(final HintManager hintManager) {
        Preconditions.checkState(null == HINT_MANAGER_HOLDER.get(), "HintManagerHolder has previous value, please clear first.");
        HINT_MANAGER_HOLDER.set(hintManager);
    }

    public static boolean isUseShardingHint() {
        // 判斷當前線程中是否使用了sharding hint--即HintManager中的shardingHint爲true
        return null != HINT_MANAGER_HOLDER.get() && HINT_MANAGER_HOLDER.get().isShardingHint();
    }

    public static Optional<ShardingValue<?>> getDatabaseShardingValue(final ShardingKey shardingKey) {
        // 如果使用了sharding hint，那麼從ThreadLocal中取數據庫的sharding值
        return isUseShardingHint() ? Optional.<ShardingValue<?>>fromNullable(HINT_MANAGER_HOLDER.get().getDatabaseShardingValue(shardingKey)) : Optional.<ShardingValue<?>>absent();
    }

    public static Optional<ShardingValue<?>> getTableShardingValue(final ShardingKey shardingKey) {
        // 如果使用了sharding hint，那麼從ThreadLocal中取表的sharding值
        return isUseShardingHint() ? Optional.<ShardingValue<?>>fromNullable(HINT_MANAGER_HOLDER.get().getTableShardingValue(shardingKey)) : Optional.<ShardingValue<?>>absent();
    }

    public static boolean isMasterRouteOnly() {
        // 是否強制路由主庫--sharding-jdbc的特性之一：強制路由
        return null != HINT_MANAGER_HOLDER.get() && HINT_MANAGER_HOLDER.get().isMasterRouteOnly();
    }

    public static boolean isDatabaseShardingOnly() {
        // 是否只是數據庫sharding
        return null != HINT_MANAGER_HOLDER.get() && HINT_MANAGER_HOLDER.get().isDatabaseShardingOnly();
    }

    /**
     * Clear hint manager for current thread-local.
     */
    public static void clear() {
        // ThreadLocal用完需要清理
        HINT_MANAGER_HOLDER.remove();
    }
}

ThreadLocal中管理的HintManager定義如下：

@NoArgsConstructor(access = AccessLevel.PRIVATE)
public final class HintManager implements AutoCloseable {

    // 數據庫強制路由的值
    private final Map<ShardingKey, ShardingValue<?>> databaseShardingValues = new HashMap<>();

    // 表強制路由的值
    private final Map<ShardingKey, ShardingValue<?>> tableShardingValues = new HashMap<>();

    // 即是否使用了強制路由特性
    @Getter
    private boolean shardingHint;

    // 是否強制路由到主數據庫
    @Getter
    private boolean masterRouteOnly;

    @Getter
    private boolean databaseShardingOnly;

    ... ...

    @Override
    public void close() {
        HintManagerHolder.clear();
    }
}

sharding值保存在ThreadLocal中，所以需要在操作結束時調用HintManager.close()來清除ThreadLocal中的內容。HintManager實現了AutoCloseable接口，推薦使用try with resource（JDK7新特性，參考Java 7中的Try-with-resources）自動關閉清理ThreadLocl線程中的數據。

如何使用

分析了sharding-jdbc的強制路由實現的源碼，接下來說說如何使用這一niubility特性，假定數據源定義如下：

private static ShardingDataSource getShardingDataSource() throws SQLException {
    DataSourceRule dataSourceRule = new DataSourceRule(createDataSourceMap());
    TableRule orderTableRule = TableRule
            .builder("t_order")
            .actualTables(Arrays.asList("t_order_0", "t_order_1"))
            .dataSourceRule(dataSourceRule)
            .build();
    TableRule orderItemTableRule = TableRule
            .builder("t_order_item")
            .actualTables(Arrays.asList("t_order_item_0", "t_order_item_1"))
            .dataSourceRule(dataSourceRule)
            .build();
    ShardingRule shardingRule = ShardingRule.builder()
            .dataSourceRule(dataSourceRule)
            .tableRules(Arrays.asList(orderTableRule, orderItemTableRule))
            .bindingTableRules(Collections.singletonList(new BindingTableRule(Arrays.asList(orderTableRule, orderItemTableRule))))
            .databaseShardingStrategy(new DatabaseShardingStrategy("user_id", new ModuloDatabaseShardingAlgorithm()))
            .tableShardingStrategy(new TableShardingStrategy("order_id", new ModuloTableShardingAlgorithm())).build();
    return new ShardingDataSource(shardingRule);
}

根據數據源定義可知，數據庫的sharding column爲user_id，表的sharding column爲order_id；

1、強制路由數據庫

• 如何使用

private static void printHintSimpleSelect(final DataSource dataSource) throws SQLException {
    // SQL語句並不涉及任何數據庫路由和表路由信息（即where語句中沒有user_id條件和order_id條件）
    String sql = "SELECT i.* FROM t_order o JOIN t_order_item i ON o.order_id=i.order_id";
    try (
            HintManager hintManager = HintManager.getInstance();
            Connection conn = dataSource.getConnection();
            PreparedStatement preparedStatement = conn.prepareStatement(sql)) {
        // 強制路由：數據庫路由sharding column即user_id的值爲10
        hintManager.addDatabaseShardingValue("t_order", "user_id", 10);
        try (ResultSet rs = preparedStatement.executeQuery()) {
            //todo do something
        }
    }
}

由於指定了強制路由數據庫的值user_id=10，所以只會輸出ds_jdbc_0這個庫中符合條件的數據。而ds_jdbc_1會被過濾；

• 實現原理

private Collection<String> routeDataSources(final TableRule tableRule) {
    // 首先得到數據庫sharding策略，例如：數據庫按照列user_id進行sharding
    DatabaseShardingStrategy strategy = shardingRule.getDatabaseShardingStrategy(tableRule);
    // 然後從ThreadLocal中取出sharding的值
    List<ShardingValue<?>> shardingValues = HintManagerHolder.isUseShardingHint() ? getDatabaseShardingValuesFromHint(strategy.getShardingColumns())
            : getShardingValues(strategy.getShardingColumns());
    Collection<String> result = strategy.doStaticSharding(tableRule.getActualDatasourceNames(), shardingValues);
    Preconditions.checkState(!result.isEmpty(), "no database route info");
    return result;
}

2、強制路由表

• 如何使用

private static void printHintSimpleSelect(final DataSource dataSource) throws SQLException {
    // SQL語句並不涉及任何數據庫路由和表路由信息（即where語句中沒有user_id條件和order_id條件）
    String sql = "SELECT i.* FROM t_order o JOIN t_order_item i ON o.order_id=i.order_id";
    try (
            HintManager hintManager = HintManager.getInstance();
            Connection conn = dataSource.getConnection();
            PreparedStatement preparedStatement = conn.prepareStatement(sql)) {
        // 強制路由：表路由sharding column即order_id的值爲1000
        hintManager.addTableShardingValue("t_order", "order_id", 1000);
        try (ResultSet rs = preparedStatement.executeQuery()) {
            //todo do something
        }
    }
}

由於指定了強制路由表的值order_id=1000，所以只會輸出所有庫中與t_order_0 匹配的數據。而與t_order_1匹配的數據會被過濾；

• 實現原理

private Collection<String> routeTables(final TableRule tableRule, final String routedDataSource) {
    // 首先得到表的sharding策略，例如：表按照列order_id進行sharding
    TableShardingStrategy strategy = shardingRule.getTableShardingStrategy(tableRule);
    // 然後從ThreadLocal中取出sharding的值
    List<ShardingValue<?>> shardingValues = HintManagerHolder.isUseShardingHint() ? getTableShardingValuesFromHint(strategy.getShardingColumns())
            : getShardingValues(strategy.getShardingColumns());
    Collection<String> result = tableRule.isDynamic() ? strategy.doDynamicSharding(shardingValues) : strategy.doStaticSharding(tableRule.getActualTableNames(routedDataSource), shardingValues);
    Preconditions.checkState(!result.isEmpty(), "no table route info");
    return result;
}

3、強制路由主庫

• 如何使用

HintManager hintManager = HintManager.getInstance();
hintManager.setMasterRouteOnly();

• 實現原理

public NamedDataSource getDataSource(final SQLType sqlType) {
    // 是否滿足主庫路由的條件
    if (isMasterRoute(sqlType)) {
        // 路由主庫後，ThreadLocal方式設置DML_FLAG爲true，從而剩餘的SQL全部路由主庫
        DML_FLAG.set(true);
        // 直接根據主數據源名稱返回對應的數據源
        return new NamedDataSource(masterDataSourceName, masterDataSource);
    }

    // 如果沒有滿足主庫路由的條件，那麼根據負載均衡策略選擇一個從庫（默認輪詢策略）
    String selectedSourceName = masterSlaveLoadBalanceStrategy.getDataSource(name, masterDataSourceName, new ArrayList<>(slaveDataSources.keySet()));
    DataSource selectedSource = selectedSourceName.equals(masterDataSourceName) ? masterDataSource : slaveDataSources.get(selectedSourceName);
    Preconditions.checkNotNull(selectedSource, "");
    return new NamedDataSource(selectedSourceName, selectedSource);
}

private boolean isMasterRoute(final SQLType sqlType) {
    return SQLType.DQL != sqlType || DML_FLAG.get() || HintManagerHolder.isMasterRouteOnly();
}

主從數據源配置：

<rdb:master-slave-data-source id="dbtbl_0" master-data-source-ref="dbtbl_0_master" slave-data-sources-ref="dbtbl_0_slave_0, dbtbl_0_slave_1" />
<rdb:master-slave-data-source id="dbtbl_1" master-data-source-ref="dbtbl_1_master" slave-data-sources-ref="dbtbl_1_slave_0, dbtbl_1_slave_1" />

從isMasterRoute(SQLType)方法可知，主庫路由的條件如下：
1. 非DQL語句即可（即DML和DDL強制主庫路由）；
2. 批量執行的SQL中已經有主庫路由後，剩餘的SQL都路由主庫；
3. hintManager.setMasterRouteOnly()強行路由主庫；