Hibernate统计日志记录

Question

我有点卡住解释hibernate会话日志。我的主要问题是许多查询都很慢 - 基于我实现的一些TimeWatch记录。为了进一步追踪问题，我已经启用了hibernate会话日志记录，目的是查看执行查询或获取连接的时间是否丢失（这意味着错误配置，我猜）。

关于用例的一点点 - Oracle DB，Spring，Hibernate。在“繁忙时期”，有一个最大值。 15个线程对数据库执行查询。所以没什么特别的，我猜。

现在我看到hibernate会话日志就像。

2017-03-30 13:35:13.834+0200 [process-documents-task-6] I [/] o.h.e.i.StatisticalLoggingSessionEventListener - Session Metrics {
    636713687 nanoseconds spent acquiring 1 JDBC connections;
    57993 nanoseconds spent releasing 1 JDBC connections;
    636859879 nanoseconds spent preparing 1 JDBC statements;
    2231526 nanoseconds spent executing 1 JDBC statements;
    0 nanoseconds spent executing 0 JDBC batches;
    0 nanoseconds spent performing 0 L2C puts;
    0 nanoseconds spent performing 0 L2C hits;
    0 nanoseconds spent performing 0 L2C misses;
    0 nanoseconds spent executing 0 flushes (flushing a total of 0 entities and 0 collections);
    9261 nanoseconds spent executing 1 partial-flushes (flushing a total of 0 entities and 0 collections)

要么

2017-03-30 13:35:16.073+0200 [process-documents-task-8] I [/] o.h.e.i.StatisticalLoggingSessionEventListener - Session Metrics {
    2893793341 nanoseconds spent acquiring 1 JDBC connections;
    22196 nanoseconds spent releasing 1 JDBC connections;
    2893869403 nanoseconds spent preparing 1 JDBC statements;
    1509926 nanoseconds spent executing 1 JDBC statements;
    0 nanoseconds spent executing 0 JDBC batches;
    0 nanoseconds spent performing 0 L2C puts;
    0 nanoseconds spent performing 0 L2C hits;
    0 nanoseconds spent performing 0 L2C misses;
    0 nanoseconds spent executing 0 flushes (flushing a total of 0 entities and 0 collections);
    4056 nanoseconds spent executing 1 partial-flushes (flushing a total of 0 entities and 0 collections)

但究竟是什么意思呢？

我现在解释的是什么

• 执行查询的时间为1509926纳秒（1.5毫秒）：似乎没问题
• 2893793341纳秒获取连接（2.8秒）：不行:(
• 准备声明2893869403纳秒（2.8秒）：不行:(

准备声明是什么意思？根据我从javadoc中读取的内容，它可能意味着将查询发送到数据库进行优化。但为什么它总是在获取连接需求的时候呢？

关于如何追查问题的任何建议，根本原因更进一步？

或任何有关问题的提示？

谢谢你的帮助！

最好的问候，Stefan

Answer 1

默认的Hibernate统计信息不是很有用。我们计划在未来的Hibernate版本中增强这一方面。

但是，正如我在my book中所解释的那样，您可以提供基于Statistics构建的自己的Dropwizard Metrics实现。

简而言之，假设你有一个扩展TransactionStatistics的org.hibernate.stat.internal.ConcurrentStatisticsImpl类：

public class TransactionStatistics extends ConcurrentStatisticsImpl {

    private static final ThreadLocal<AtomicLong> startNanos = new ThreadLocal<AtomicLong>() {
        @Override protected AtomicLong initialValue() {
            return new AtomicLong();
        }
    };

    private static final ThreadLocal<AtomicLong> connectionCounter = new ThreadLocal<AtomicLong>() {
        @Override protected AtomicLong initialValue() {
            return new AtomicLong();
        }
    };

    private StatisticsReport report = new StatisticsReport();

    @Override public void connect() {
        connectionCounter.get().incrementAndGet();
        startNanos.get().compareAndSet(0, System.nanoTime());
        super.connect();
    }

    @Override public void endTransaction(boolean success) {
        try {
            report.transactionTime(System.nanoTime() - startNanos.get().get());
            report.connectionsCount(connectionCounter.get().get());
            report.generate();
        } finally {
            startNanos.remove();
            connectionCounter.remove();
        }
        super.endTransaction(success);
    }
}

您需要创建一个StatisticsFactory实现：

public class TransactionStatisticsFactory implements StatisticsFactory {

    @Override
    public StatisticsImplementor buildStatistics(SessionFactoryImplementor sessionFactory) {
        return new TransactionStatistics();
    }
}

并配置如下：

就是这样！

现在，您可以监视统计数据并将其导出为Dropwizard Metrics支持的任何格式。而且，Dropwizard Metrics使用各种水库，因此您可以选择最适合您的用例。

要查看Dropwizard指标的真正力量，请查看FlexyPool。