Atomikos Forum

We have had some problems with Atomikos in combination with the 3.00.JC3 driver of Informix (the recommended version!). To provide the necessary context:
- our database restarts every night
- we add dynamically datasources (even during a transaction)
- xares_.isSameRM ( xares_ ) doesn't throw any Exception on a 'corrupt' XAResource
- Atomikos starts a recovery scan each time a new datasource is initialized

Given this we get troubles "caused" by line 784 XATransactionalResource#recover(Participant participant)

if ( recovered || getName().equals ( xarestx.resourcename_ ) )
                xarestx.setRecoveredXAResource ( getXAResource () );

getXAResource() doesn't detect the corrupt state of the XAResource and so a corrupt XAResource enters into the tranaction. This gives us following errors:

javax.transaction.RollbackException: Prepare: NO vote
    at com.atomikos.icatch.jta.TransactionImp.commit(TransactionImp.java:311)
    at com.atomikos.icatch.jta.TransactionManagerImp.commit(TransactionManagerImp.java:608)
    at com.atomikos.icatch.jta.UserTransactionImp.commit(UserTransactionImp.java:168)
    at org.springframework.transaction.jta.JtaTransactionManager.doCommit(JtaTransactionManager.java:1028)
    ... 69 more

and...

javax.transaction.xa.XAException: java.sql.SQLException: System or internal error java.net.SocketException: Broken pipe
        at com.informix.jdbcx.IfxXAResource.prepare(IfxXAResource.java:248)
        at com.atomikos.datasource.xa.XAResourceTransaction.prepare(XAResourceTransaction.java:799)
        at com.atomikos.icatch.imp.PrepareMessage.send(PrepareMessage.java:79)
        at com.atomikos.icatch.imp.PropagationMessage.submit(PropagationMessage.java:86)
        at com.atomikos.icatch.imp.Propagator$PropagatorThread.run(Propagator.java:62)
        at java.util.concurrent.ThreadPoolExecutor$Worker.runTask(Unknown Source)
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(Unknown Source)
        at java.lang.Thread.run(Unknown Source)

Our workaroud is to use a tweaked Informix driver. We changed the behaviour of isSameRM :

boolean returnValue = this.xares.isSameRM(((TweakedXAResource) xares).getXares());
if (this == xares)
{
  try
  {
    forget(new IfxXid(777, new byte[]{0x27, 0x06}, new byte[]{0x19, 0x79}));
  }
  catch (XAException e)
  {
    if (e.errorCode != XAException.XAER_NOTA)
    {
      throw e;
    }
  }
}
return returnValue;

This workaround could be avoided if Atomikos would use a simular trick to test the state of the connection:

instead of

boolean ret = true;

// check if connection has not timed out
try {
    // we should be the same as ourselves!
    // NOTE: xares_ is null if no connection could be gotten
    // in that case we just return true
    // otherwise, test the xaresource liveness
    if ( xares_ != null ) {
    xares_.isSameRM ( xares_ );
    ret = false;
    }
} catch ( XAException xa ) {
    // timed out?
    Configuration.logDebug ( servername_
        + ": XAResource needs refresh?", xa );

}
return ret;

you could...

boolean ret = true;

// check if connection has not timed out
try {
    // we should be the same as ourselves!
    // NOTE: xares_ is null if no connection could be gotten
    // in that case we just return true
    // otherwise, test the xaresource liveness
    if ( xares_ != null ) {
    xares_.forget(new IfxXid(777, new byte[]{0x27, 0x06}, new byte[]{0x19, 0x79}));
    ret = false; // should be impossible !!
    }
} catch ( XAException xa ) {
    if (e.errorCode == XAException.XAER_NOTA)
    {
      ret = false;
      // timed out?
      Configuration.logDebug ( servername_
        + ": XAResource needs refresh?", xa );
    }
}
return ret;

You will just need to alter the Xid constructor.

Is this a good sollution or are better sollutions out there?

Hi,

Thanks for the feedback. The forget trick might work for Informix but let me check if it is safe for other DBMS too. I have scheduled an issue for this to be looked at by our engineers.

Best