Receiving a Hikari Leak Detection using springs AbstractRoutingDataSource

[Warren Goldman]

The routing data source has multiple data sources configured to it. The exception below happens once in awhile, and actually somewhat infrequently (even though this call is made numerous times).

We are configuring HikariDataSource's for the AbstractRoutingDataSource.targetDataSources. The AbstractRoutingDataSource implementation is passed as a dataSource to org.springframework.orm.hibernate4.LocalSessionFactoryBean.

The below is triggered by a call to begin a transaction that was created via sessionFactory.getSession().getTransaction().begin() <-- somewhat pseudocode...

 com.zaxxer.hikari.proxy.LeakTask - Connection leak detection triggered, stack trace follows

java.lang.Exception

        at org.springframework.jdbc.datasource.lookup.AbstractRoutingDataSource.getConnection(AbstractRoutingDataSource.java:164)

        at org.hibernate.engine.jdbc.connections.internal.DatasourceConnectionProviderImpl.getConnection(DatasourceConnectionProviderImpl.java:139)

        at org.hibernate.internal.AbstractSessionImpl$NonContextualJdbcConnectionAccess.obtainConnection(AbstractSessionImpl.java:380)

        at org.hibernate.engine.jdbc.internal.LogicalConnectionImpl.obtainConnection(LogicalConnectionImpl.java:228)

        at org.hibernate.engine.jdbc.internal.LogicalConnectionImpl.getConnection(LogicalConnectionImpl.java:171)

        at org.hibernate.engine.transaction.internal.jdbc.JdbcTransaction.doBegin(JdbcTransaction.java:67)

        at org.hibernate.engine.transaction.spi.AbstractTransactionImpl.begin(AbstractTransactionImpl.java:162)

[Brett Wooldridge]

What version of HikariCP? What is the leakDetectionThreshold set to?

-Brett

[Warren Goldman]

2.4.1

<property name="leakDetectionThreshold"    value="60000"/>

[Brett Wooldridge]

If you upgrade to HikariCP 2.5.1 there is a new log that appears if a connection is "unleaked", i.e. it is eventually returned to the pool.  This can help you distinguish between a true leak and a transient spike in cpu/network/database load that pushes a query over the reporting threshold.

-Brett