Subject Re: Another backup error
Author nagypapi
I'm just hitting myself, because I found out the cause of the error:
Do not use relative paths when setting Backup path, use absolute paths

Anyway, either a more informative exception or the possibility to use
relative paths might help the next poor chap :)


--- In, "Roman Rokytskyy"
<rrokytskyy@a...> wrote:
> > Yes, both of my programs have created an instance of FBBackupManager.
> > I would be happy to create an instance only when I'm doing a backup,
> > but found no explicit way to close the FBBackupManager.
> It is automatically closed after finishing backup. There is one
issue with
> Services API. It should be considered as a job that you start on the
> and simply redirect the output to your client application. So it
really does
> not matter whether you close your application or not. If I remember
> correctly, API provides means to disconnect, then to reconnect,
> running jobs, attach to their output and so on.
> > When does FBBackupManager connect to the db?
> > I'd guess during FBBackupManager.backup(), because the username,
> > database and passwords get set separatly. Does the backupmanager close
> > the connection, when it finishes backing up?
> It should, if not - then it's a bug. The code is simple:
> IscSvcHandle svcHandle = attachServiceManager(gds);
> try {
> gds.iscServiceStart(svcHandle, srb);
> queueService(gds, svcHandle);
> } finally {
> detachServiceManager(gds, svcHandle);
> }
> The task is specified in srb - ServiceRequestBuffer, which tells the
> what to execute (backup, restore, paths to database, etc). In theory
one can
> skip the queueService operation, though I did not check how to
> later.
> > Then the other backupmanager shouldn't bother it,
> If backup is finished - it should work correctly.
> > or maybe running in the same jvm causes the problem? (both are run
> > netbeans).
> Nope, should not depend on runtime environment.
> > I'll do some tests maybe tomorrow, but in the long run I want to do
> > backups with multiple clients (from different computers) connected to
> > the same database.
> Thanks! If you happen to get a reproducable test case, please send
it to me.
> I will have a look somewhere in January.
> Roman