別再亂升級 MySQL 驅動了。。

來源:https://blog.csdn.net/qq_40378034/article/details/117851207

現象

最近同事發現新服務用的驅動都是 MySQL8.0,而一些老系統 MySQL 驅動版本較低,存在一些危險漏洞,於是主動的把一些老應用的 MySQL 驅動程序升級到了 8.0。

然而升級後,在併發量較高時,查看監控打點,Druid 連接池拿到連接並執行 SQL 的時間大部分都超過 200ms。

本文詳細的分析了這次“破案”的全過程。

對系統進行壓測,發現出現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)

通過上面的信息得知,有線程 BLOCKED 了,BLOCKED 的位置在 com.alibaba.druid.util.Utils.loadClass(Utils.java:220),於是重點的查看這一塊的代碼,最終發現了問題。

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 中類名改爲com.mysql.cj.jdbc.ConnectionImpl,所以 MySQL 驅動8.0 中加載不到com.mysql.jdbc.MySQLConnection

getLastPacketReceivedTimeMs()方法實現中,如果Utils.loadClass("com.mysql.jdbc.MySQLConnection")加載不到類並拋出異常,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,並不會拋出異常,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);
  }
 }

這是因爲 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,所以會出現線程阻塞,性能下降的現象。

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,這取決於Druid檢測連接是否有效的參數

「Druid檢測連接是否有效的參數」

  • testOnBorrow:每次獲取連接時執行validationQuery檢測連接是否有效(會影響性能)
  • testOnReturn:每次歸還連接時執行validationQuery檢測連接是否有效(會影響性能)
  • testWhileIdle:申請連接的時候檢測,如果空閒時間大於timeBetweenEvictionRunsMillis,執行validationQuery檢測連接是否有效

應用中設置了testOnBorrow=true,每次獲取連接時,都會去搶佔synchronized鎖,所以性能下降的很明顯

解決方案

經驗證,使用Druid 1.x版本<=1.1.22會出現該bug,解決方案就是升級至Druid 1.x版本>=1.1.23或者Druid 1.2.x版本

GitHub issue:https://github.com/alibaba/druid/issues/3808

近期熱文推薦:

1.1,000+ 道 Java面試題及答案整理(2021最新版)

2.別在再滿屏的 if/ else 了,試試策略模式,真香!!

3.臥槽!Java 中的 xx ≠ null 是什麼新語法?

4.Spring Boot 2.5 重磅發佈,黑暗模式太炸了!

5.《Java開發手冊(嵩山版)》最新發布,速速下載!

覺得不錯,別忘了隨手點贊+轉發哦!

發表評論
所有評論
還沒有人評論,想成為第一個評論的人麼? 請在上方評論欄輸入並且點擊發布.
相關文章