低版本Druid連接池+MySQL驅動8.0導致線程阻塞、性能受限

更新時間：2021年07月01日 15:25:13 作者：邋遢的流浪劍客

應用升級MySQL驅動8.0后，在并發(fā)量較高時，查看監(jiān)控打點，Druid連接池拿到連接并執(zhí)行SQL的時間大部分都超過200ms，本文就解決一下這個問題

現(xiàn)象

應用升級MySQL驅動8.0后，在并發(fā)量較高時，查看監(jiān)控打點，Druid連接池拿到連接并執(zhí)行SQL的時間大部分都超過200ms

對系統(tǒng)進行壓測，發(fā)現(xiàn)出現(xiàn)大量線程阻塞的情況，線程dump信息如下：

"http-nio-5366-exec-48" #210 daemon prio=5 os_prio=0 tid=0x00000000023d0800 nid=0x3be9 waiting for monitor entry [0x00007fa4c1400000]
   java.lang.Thread.State: BLOCKED (on object monitor)
        at org.springframework.boot.web.embedded.tomcat.TomcatEmbeddedWebappClassLoader.loadClass(TomcatEmbeddedWebappClassLoader.java:66)
        - waiting to lock <0x0000000775af0960> (a java.lang.Object)
        at org.apache.catalina.loader.WebappClassLoaderBase.loadClass(WebappClassLoaderBase.java:1186)
        at com.alibaba.druid.util.Utils.loadClass(Utils.java:220)
        at com.alibaba.druid.util.MySqlUtils.getLastPacketReceivedTimeMs(MySqlUtils.java:372)

根因分析

public class MySqlUtils {

    public static long getLastPacketReceivedTimeMs(Connection conn) throws SQLException {
        if (class_connectionImpl == null && !class_connectionImpl_Error) {
            try {
                class_connectionImpl = Utils.loadClass("com.mysql.jdbc.MySQLConnection");
            } catch (Throwable error){
                class_connectionImpl_Error = true;
            }
        }

        if (class_connectionImpl == null) {
            return -1;
        }

        if (method_getIO == null && !method_getIO_error) {
            try {
                method_getIO = class_connectionImpl.getMethod("getIO");
            } catch (Throwable error){
                method_getIO_error = true;
            }
        }

        if (method_getIO == null) {
            return -1;
        }

        if (class_MysqlIO == null && !class_MysqlIO_Error) {
            try {
                class_MysqlIO = Utils.loadClass("com.mysql.jdbc.MysqlIO");
            } catch (Throwable error){
                class_MysqlIO_Error = true;
            }
        }

        if (class_MysqlIO == null) {
            return -1;
        }

        if (method_getLastPacketReceivedTimeMs == null && !method_getLastPacketReceivedTimeMs_error) {
            try {
                Method method = class_MysqlIO.getDeclaredMethod("getLastPacketReceivedTimeMs");
                method.setAccessible(true);
                method_getLastPacketReceivedTimeMs = method;
            } catch (Throwable error){
                method_getLastPacketReceivedTimeMs_error = true;
            }
        }

        if (method_getLastPacketReceivedTimeMs == null) {
            return -1;
        }

        try {
            Object connImpl = conn.unwrap(class_connectionImpl);
            if (connImpl == null) {
                return -1;
            }

            Object mysqlio = method_getIO.invoke(connImpl);
            Long ms = (Long) method_getLastPacketReceivedTimeMs.invoke(mysqlio);
            return ms.longValue();
        } catch (IllegalArgumentException e) {
            throw new SQLException("getLastPacketReceivedTimeMs error", e);
        } catch (IllegalAccessException e) {
            throw new SQLException("getLastPacketReceivedTimeMs error", e);
        } catch (InvocationTargetException e) {
            throw new SQLException("getLastPacketReceivedTimeMs error", e);
        }
    }

MySqlUtils中的getLastPacketReceivedTimeMs()方法會加載com.mysql.jdbc.MySQLConnection這個類，但在MySQL驅動8.0中類名改為com.mysql.cj.jdbc.ConnectionImpl，所以MySQL驅動8.0中加載不到com.mysql.jdbc.MySQLConnection

getLastPacketReceivedTimeMs()方法實現(xiàn)中，如果Utils.loadClass("com.mysql.jdbc.MySQLConnection")加載不到類并拋出異常，會修改變量class_connectionImpl_Error，下次調用不會再進行加載

public class Utils {

    public static Class<?> loadClass(String className) {
        Class<?> clazz = null;

        if (className == null) {
            return null;
        }

        try {
            return Class.forName(className);
        } catch (ClassNotFoundException e) {
            // skip
        }

        ClassLoader ctxClassLoader = Thread.currentThread().getContextClassLoader();
        if (ctxClassLoader != null) {
            try {
                clazz = ctxClassLoader.loadClass(className);
            } catch (ClassNotFoundException e) {
                // skip
            }
        }

        return clazz;
    }

但是，在Utils的loadClass()方法中同樣catch了ClassNotFoundException，這就導致loadClass()在加載不到類的時候，并不會拋出異常，從而會導致每調用一次getLastPacketReceivedTimeMs()方法，就會加載一次MySQLConnection這個類

線程dump信息中可以看到是在調用TomcatEmbeddedWebappClassLoader的loadClass()方法時，導致線程阻塞的

public class TomcatEmbeddedWebappClassLoader extends ParallelWebappClassLoader {

 public Class<?> loadClass(String name, boolean resolve) throws ClassNotFoundException {
  synchronized (JreCompat.isGraalAvailable() ? this : getClassLoadingLock(name)) {
   Class<?> result = findExistingLoadedClass(name);
   result = (result != null) ? result : doLoadClass(name);
   if (result == null) {
    throw new ClassNotFoundException(name);
   }
   return resolveIfNecessary(result, resolve);
  }
 }

這是因為TomcatEmbeddedWebappClassLoader在加載類的時候，會加synchronized鎖，這就導致每調用一次getLastPacketReceivedTimeMs()方法，就會加載一次com.mysql.jdbc.MySQLConnection，而又始終加載不到，在加載類的時候會加synchronized鎖，所以會出現(xiàn)線程阻塞，性能下降的現(xiàn)象

getLastPacketReceivedTimeMs()方法調用時機

public abstract class DruidAbstractDataSource extends WrapperAdapter implements DruidAbstractDataSourceMBean, DataSource, DataSourceProxy, Serializable {

    protected boolean testConnectionInternal(DruidConnectionHolder holder, Connection conn) {
        String sqlFile = JdbcSqlStat.getContextSqlFile();
        String sqlName = JdbcSqlStat.getContextSqlName();

        if (sqlFile != null) {
            JdbcSqlStat.setContextSqlFile(null);
        }
        if (sqlName != null) {
            JdbcSqlStat.setContextSqlName(null);
        }
        try {
            if (validConnectionChecker != null) {
                boolean valid = validConnectionChecker.isValidConnection(conn, validationQuery, validationQueryTimeout);
                long currentTimeMillis = System.currentTimeMillis();
                if (holder != null) {
                    holder.lastValidTimeMillis = currentTimeMillis;
                    holder.lastExecTimeMillis = currentTimeMillis;
                }

                if (valid && isMySql) { // unexcepted branch
                    long lastPacketReceivedTimeMs = MySqlUtils.getLastPacketReceivedTimeMs(conn);
                    if (lastPacketReceivedTimeMs > 0) {
                        long mysqlIdleMillis = currentTimeMillis - lastPacketReceivedTimeMs;
                        if (lastPacketReceivedTimeMs > 0 //
                                && mysqlIdleMillis >= timeBetweenEvictionRunsMillis) {
                            discardConnection(holder);
                            String errorMsg = "discard long time none received connection. "
                                    + ", jdbcUrl : " + jdbcUrl
                                    + ", jdbcUrl : " + jdbcUrl
                                    + ", lastPacketReceivedIdleMillis : " + mysqlIdleMillis;
                            LOG.error(errorMsg);
                            return false;
                        }
                    }
                }

                if (valid && onFatalError) {
                    lock.lock();
                    try {
                        if (onFatalError) {
                            onFatalError = false;
                        }
                    } finally {
                        lock.unlock();
                    }
                }

                return valid;
            }

            if (conn.isClosed()) {
                return false;
            }

            if (null == validationQuery) {
                return true;
            }

            Statement stmt = null;
            ResultSet rset = null;
            try {
                stmt = conn.createStatement();
                if (getValidationQueryTimeout() > 0) {
                    stmt.setQueryTimeout(validationQueryTimeout);
                }
                rset = stmt.executeQuery(validationQuery);
                if (!rset.next()) {
                    return false;
                }
            } finally {
                JdbcUtils.close(rset);
                JdbcUtils.close(stmt);
            }

            if (onFatalError) {
                lock.lock();
                try {
                    if (onFatalError) {
                        onFatalError = false;
                    }
                } finally {
                    lock.unlock();
                }
            }

            return true;
        } catch (Throwable ex) {
            // skip
            return false;
        } finally {
            if (sqlFile != null) {
                JdbcSqlStat.setContextSqlFile(sqlFile);
            }
            if (sqlName != null) {
                JdbcSqlStat.setContextSqlName(sqlName);
            }
        }
    }

只有DruidAbstractDataSource的testConnectionInternal()方法中會調用getLastPacketReceivedTimeMs()方法

testConnectionInternal()是用來檢測連接是否有效的，在獲取連接和歸還連接時都有可能會調用該方法，這取決于Druid檢測連接是否有效的參數(shù)

Druid檢測連接是否有效的參數(shù)：

testOnBorrow：每次獲取連接時執(zhí)行validationQuery檢測連接是否有效（會影響性能）
testOnReturn：每次歸還連接時執(zhí)行validationQuery檢測連接是否有效（會影響性能）
testWhileIdle：申請連接的時候檢測，如果空閑時間大于timeBetweenEvictionRunsMillis，執(zhí)行validationQuery檢測連接是否有效
應用中設置了testOnBorrow=true，每次獲取連接時，都會去搶占synchronized鎖，所以性能下降的很明顯