背景介绍
在一次druid数据库连接池连接泄露的排查分析介绍了连接泄露的分析排查过程，在几行代码轻松复现druid连接泄露的BUG之PhyTimeout介绍了当配置了phyTimeoutMillis参数情况下，连接泄露的场景，在几行代码轻松复现druid连接泄露的BUG之onFatalError介绍了当数据库操作出现某些异常情况下，连接泄露的场景，下面通过代码的方式来复现当配置了keepalive选项的情况下出现连接泄露的场景。
复现过程
连接泄露场景
模拟当配置了keepalive选项的情况下出现连接泄露的场景。
连接泄露模拟代码
import com.alibaba.druid.pool.DruidConnectionHolder; import com.alibaba.druid.pool.DruidDataSource; import com.alibaba.druid.pool.DruidPooledConnection; import java.lang.reflect.Field; import java.sql.SQLException; import java.util.HashMap; import java.util.List; import java.util.Map; public class DruidAbandonedCase4Keepalive { public static void main(String[] args) throws Exception { Map<Long,DruidConnectionHolder> holderMap = new HashMap<>(); DruidDataSource dataSource = new DruidDataSource(); dataSource.setDriverClassName("com.mysql.cj.jdbc.Driver"); dataSource.setUsername("root"); dataSource.setPassword("123456"); dataSource.setUrl("jdbc:mysql://127.0.0.1:3306/test?serverTimezone=UTC&useUnicode=true&characterEncoding=utf-8&useSSL=false"); dataSource.setMinIdle(2); dataSource.setKeepAlive(true); dataSource.setTimeBetweenEvictionRunsMillis(500); long minEvictableIdleTimeMillis = 500L; dataSource.setMinEvictableIdleTimeMillis(minEvictableIdleTimeMillis); long keepAliveBetweenTimeMillis = 1000L; dataSource.setKeepAliveBetweenTimeMillis(keepAliveBetweenTimeMillis); dataSource.init(); try{ Field destroyConnectionThreadField = DruidDataSource.class.getDeclaredField("destroyConnectionThread"); destroyConnectionThreadField.setAccessible(true); Thread destroyConnectionThread = (Thread)destroyConnectionThreadField.get(dataSource); destroyConnectionThread.interrupt(); Thread.State state = destroyConnectionThread.getState(); System.out.println("destroyConnectionThread state : " + state); }catch (Exception e){ e.printStackTrace(); } DruidPooledConnection connection1 = dataSource.getConnection(); DruidConnectionHolder holder1 = connection1.getConnectionHolder(); print(holder1); holderMap.put(holder1.getConnectionId(),holder1); DruidPooledConnection connection2 = dataSource.getConnection(); DruidConnectionHolder holder2 = connection2.getConnectionHolder(); print(holder2); holderMap.put(holder2.getConnectionId(),holder2); connection2.close(); sleep(keepAliveBetweenTimeMillis - minEvictableIdleTimeMillis + 100L); connection1.close(); sleep(minEvictableIdleTimeMillis - 100L); pooling(dataSource,holderMap); shrink(dataSource); pooling(dataSource,holderMap); sleep(100L); shrink(dataSource); pooling(dataSource,holderMap); System.out.println(); print(holder1); print(holder2); } private static void sleep(long millis){ try{ Thread.sleep(millis); }catch (InterruptedException e){ e.printStackTrace(); } } private static void shrink(DruidDataSource dataSource){ System.out.println(); System.out.println("调用shrink"); // 手动执行shrink dataSource.shrink(true); } private static void pooling(DruidDataSource dataSource,Map<Long,DruidConnectionHolder> holderMap) throws SQLException { System.out.println(); System.out.println("连接池中连接情况,begin"); List<Map<String, Object>> conns = dataSource.getPoolingConnectionInfo(); for (Map<String, Object> conn : conns) { print(holderMap.get(conn.get("connectionId"))); } System.out.println("连接池中连接情况,end"); } private static void print(DruidConnectionHolder holder) throws SQLException { System.out.println(holder.getConnectionId() + /*" : " + holder +*/ " idleMillis : " + (System.currentTimeMillis() - holder.getLastActiveTimeMillis()) + " isClosed:" + holder.getConnection().isClosed()); } }
运行以上程序，输出结果如下(druid 1.2.8)：
destroyConnectionThread state : TIMED_WAITING 10001 idleMillis : 6 isClosed:false 10002 idleMillis : 1 isClosed:false 连接池中连接情况,begin 10002 idleMillis : 1015 isClosed:false 10001 idleMillis : 404 isClosed:false 连接池中连接情况,end 调用shrink 连接池中连接情况,begin 10001 idleMillis : 405 isClosed:false 10002 idleMillis : 1016 isClosed:false 连接池中连接情况,end 调用shrink 连接池中连接情况,begin 10001 idleMillis : 405 isClosed:false 10002 idleMillis : 1016 isClosed:false 连接池中连接情况,end 调用shrink 连接池中连接情况,begin 10002 idleMillis : 1133 isClosed:true 连接池中连接情况,end 10001 idleMillis : 522 isClosed:false 10002 idleMillis : 1133 isClosed:true
代码执行逻辑描述：
- 为了便于测试，中断Druid-ConnectionPool-Destroy-xx线程，以便手动调用shrink；
- 获取connectionId为1001的连接，打印连接信息；
- 获取connectionId为1002的连接，打印连接信息；
- 调用close，归还connectionId为1002的连接
- sleep (keepAliveBetweenTimeMillis - minEvictableIdleTimeMillis + 100L)时长
- 调用close，归还connectionId为1001的连接
- sleep (minEvictableIdleTimeMillis - 100L)时长（两个sleep是为了让1002 idleMillis大于keepAliveBetweenTimeMillis，让1001小于minEvictableIdleTimeMillis）
- 打印连接池中连接信息
- 执行shrink
- 打印连接池中连接信息
- sleep 100ms
- 执行shrink
- 打印连接池中连接信息
- 打印1001和1002连接信息
期望执行结果：
- connectionId为1001和1002的连接都应该被连接池管理，且连接状态都应该是未关闭状态
实际执行结果：
- connectionId为1001的连接已从连接池中删除，且连接状态是未关闭状态，泄露的连接【不符合期望】；
- connectionId为1002的连接虽然在连接池中，但是连接状态是关闭状态【不符合期望】。
测试看，keepalive选项是从druid 1.1.16引入的，druid 1.1.16-1.1.24，1.2.0-1.2.17都存在连接泄露问题，druid 1.2.18-1.2.20不存在连接泄露问题。
