RemoteExceptions and Entity bean failover

Question

The documentation gives an example of how to achieve failover with
          entity beans. (See below). How can one tell that the RemoteException
          is due to the server being down rather than some other reason.
          For example in EJB 1.0 aren't many exceptions generated in the bean
          wrapped in a RemoteException? For example if there is an SQLException
          the bean might throw a RemoteException that wraps the SQLException. In
          such a case it doesn't make sense to try it again.
          So what is the correct way to find out whether the exception is due to a
          server failure rather than some other reason?
          dan
          AccountHome home =
          (AccountHome)ctx.lookup("example.AccountHome");
          Account account;
          int attempts;
          while (attempts < LIMIT) {
          try {
          tx.begin();
          account = home.findByPrimaryKey(ACCOUNT);
          account.deposit(500);
          account.withdraw(10);
          tx.commit();
          break;
          catch (RemoteException re ) {
          // Retry
          try {
          tx.rollback();
          attempts++;
          catch (Exception e) {
          // Something unrecoverable has happened
          throw e;
          

Answer 1

Yes, this is the "nasty" case as I've heard several people put it. It's among the
          problems that make distributed programming difficult. Since the database is managing
          the transaction log, the other app servers in the cluster do not know if the tx was
          committed or not either.
          -- Rob
          dan benanav wrote:
          > If a remote exception is thrown from an entity bean then is it the case that we know
          > the transaction has rolled back? What if the server went down after the transaction
          > was committed?
          >
          > dan
          >
          > Rob Woollen wrote:
          >
          > > dan benanav wrote:
          > >
          > > > Isn't there a more definitive way to determine whether the problem is due to a
          > > > connection error or the server being down?
          > > >
          > >
          > > My first answer to your question is why do you need this? Both a db
          > > connection error and the server going down should be relatively rare events in a
          > > production system. RemoteException is meant to indicate that a system-level
          > > error has occured and the transaction has rolled back. Doing another find in
          > > this case is not going to limit your application's performance / scalability.
          > >
          > > -- Rob
          > >
          > > >
          > > > Rob Woollen wrote:
          > > >
          > > > > dan benanav wrote:
          > > > >
          > > > > > The documentation gives an example of how to achieve failover with
          > > > > > entity beans. (See below). How can one tell that the RemoteException
          > > > > > is due to the server being down rather than some other reason.
          > > > >
          > > > > Any application error should be thrown as an application exception that
          > > > > you could obviously distinguish from RemoteException.
          > > > >
          > > > > As you correctly point out, db errors like SQLException are often
          > > > > wrapped in RemoteException when they reach the client.
          > > > >
          > > > > I would code the application to give up if it gets an application
          > > > > exception. For instance, if you try to withdraw more money than you have,
          > > > > there is no sense in retrying.
          > > > >
          > > > > I would retry (up to a small limit) on RemoteExceptions if it makes
          > > > > sense in your application. Even SQLExceptions may be transient. For
          > > > > instance, Oracle may not have been able to serialize your transaction on
          > > > > the first attempt, but your second try might go through.
          > > > >
          > > > > -- Rob
          > > > >
          > > > > >
          > > > > >
          > > > > > For example in EJB 1.0 aren't many exceptions generated in the bean
          > > > > > wrapped in a RemoteException? For example if there is an SQLException
          > > > > > the bean might throw a RemoteException that wraps the SQLException. In
          > > > > > such a case it doesn't make sense to try it again.
          > > > > >
          > > > > > So what is the correct way to find out whether the exception is due to a
          > > > > > server failure rather than some other reason?
          > > > > >
          > > > > > dan
          > > > > >
          > > > > > AccountHome home =
          > > > > > (AccountHome)ctx.lookup("example.AccountHome");
          > > > > > Account account;
          > > > > > int attempts;
          > > > > >
          > > > > > while (attempts < LIMIT) {
          > > > > > try {
          > > > > > tx.begin();
          > > > > > account = home.findByPrimaryKey(ACCOUNT);
          > > > > > account.deposit(500);
          > > > > > account.withdraw(10);
          > > > > > tx.commit();
          > > > > > break;
          > > > > > }
          > > > > > catch (RemoteException re ) {
          > > > > > // Retry
          > > > > > try {
          > > > > > tx.rollback();
          > > > > > attempts++;
          > > > > > }
          > > > > > catch (Exception e) {
          > > > > > // Something unrecoverable has happened
          > > > > > throw e;
          > > > > > }
          > > > > > }