Warning in the logs about cleaning up leaked connection after upgrade from 1.3.6 to 1.3.8

[Guillaume Smet]

Hi,

After upgrading to Hikari 1.3.8, I have a lot of the following warnings in my logs:
[2014-04-28 18:35:37,107] WARN  - ThreadedHousekeeper        -  - cleaning up leaked connection
  at com.zaxxer.hikari.util.DriverDataSource.getConnection(DriverDataSource.java:63)
  at com.zaxxer.hikari.util.DriverDataSource.getConnection(DriverDataSource.java:69)
  at com.zaxxer.hikari.HikariPool.addConnection(HikariPool.java:339)
  at com.zaxxer.hikari.HikariPool.access$300(HikariPool.java:52)
  at com.zaxxer.hikari.HikariPool$1AddConnection.run(HikariPool.java:251)
  at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:471)
  at java.util.concurrent.FutureTask.run(FutureTask.java:262)
  at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
  at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
  at java.lang.Thread.run(Thread.java:744)

Note that I'm using Spring and Hibernate to deal with the connections. I'm not doing anything fancy.

Is this expected? I didn't have this issue with 1.3.6.

--
Guillaume

[Brett Wooldridge]

ThreadedHousekeeper is a class from (I think) the pgjdbc-ng driver.  Is it possible that you've updated the PostgreSQL driver?  I've told the developers over at pgjdbc-ng before that enabling their leak detection by default is likely to cause issues for connection pools.  Their leak detection triggers when a connection is created but is not closed for a long time.  However, with a connection pool this is exactly what one expects to happen; the connection is created and resides in the pool without being closed for possibly hours or days.

The changes between HikariCP 1.3.6 and 1.3.8 include only:

Fix issue #63 improper setting of null state during connection failure.

Fix isConnectionAlive() check when using Connection.isValid().

Add a JNDI factory to create datasources for Tomcat users.

Eliminate double-execution of shutdown code.

It is possible that the isConnectionAlive() method was (in 1.3.6) erroneously causing connections to close early, and therefore avoiding the pgjdbc-ng leak detection.  And the fix to isConnectionAlive() is allowing connections to live longer, and triggering the pgjdbc-ng leak detection.  However, the 1.3.7/1.3.8 behavior is correct, so I would investigate how to disable the leak detection or ThreadedHousekeeper pgjdbc-ng.  You might also consider opening a bug against pgjdbc-ng stating that their "leak" detection is causing spurious warnings for connection pools.

[Guillaume Smet]

Hi Brett,

Indeed, using housekeeper.enabled=false did the trick.

I'll wait to see if the driver development is still alive before opening a new bug. I already have a couple of pull requests waiting for attention there.

Thanks for your feedback.

--
Guillaume

[Brett Wooldridge]

I've joined as a committer on the pgjdbc-ng project, so hopefully with 3 committers on-board it can move along a little more quickly.

-Brett