Forums

by Carl Ranson >> Sun, 21 Mar 1999 22:03:50 GMT

Hi all,

Has anyone here used the warm backup code written (I think) by JUG?

The problem I have is that any beginTransaction that occurs while the backup is in progress causes an exception 3079 "Database is locked for archive backup".

I wanted to write an exception handler that displayed a progress box to the user and automagically retried the transaction when the backup had finished.

The only thing preventing this is that a 3079 is not continuable, which means the only way to deal with the exception is to do a resume_next or abort.

Anyone encountered this problem and know of a fix?
thanks,
Carl

by Craig Shearer >> Tue, 23 Mar 1999 2:15:10 GMT

Hi Carl

I have used the Lukewarm backup facility with great success however my solution to the problem you describe was simply to present a message box telling the user that the database is being backed up, and to retry the operation. In this case the Abort_Action works perfectly.

I guess that the problem that you could encounter is that you might (reasonably or unreasonably) expect that a beginTransaction instruction would never fail, and therefore have coded your application in such as way that it will break if it does - such is life, I guess

So, it would appear that the only viable solution would be to get JADE modified so that a 3079 is continuable - how hard this would be is anyone's guess.

Of course, you could have some form of semaphore object that needs to be acquired by the application before it can go into the backup operation - then you can use this to prevent it from doing so during your critical operations in the application.

You could also wait for JADE 5.0....

Craig.

by Carl Ranson >> Tue, 23 Mar 1999 3:38:08 GMT

Hi Carl

I have used the Lukewarm backup facility with great success however my solution to the problem you describe was simply to present a message box telling the user that the database is being backed up, and to retry the operation. In this case the Abort_Action works perfectly.

I guess that the problem that you could encounter is that you might (reasonably or unreasonably) expect that a beginTransaction instruction would never fail, and therefore have coded your application in such as way that it will break if it does - such is life, I guess

Such as having any chain of code that does multple transactions, for instance.
The thing that was so aggravating was that I didn't discover that the ExContinue was invalid until *after* I had done all the other work.

So, it would appear that the only viable solution would be to get JADE modified so that a 3079 is continuable - how hard this would be is anyone's guess.

I suppose one could make their own beginTransaction method that waited for the backup semaphore to clear before doing a beginTransaction and returning. This would achieve the same effect and bypass the need for an exception handler. Does add some overhead though.

I'll post a NFS to get 3079's continuable.