Newbie Questions: setabort, setcomplete, autocomplete

[Paul]

I'm a newbie and am trying to make sure I understand a few things about MTS
/ COM+ correctly. If anyone could answer the following questions I'd appreciate
it!:


1.
I understand that to take advantage of MTS or COM+'s JIT (just in time) activation
and deactivation features, you can call the setabort and setcomplete methods
of the objectcontext to deactivate the object. I've also read it's used
to control transactions ... this is the part I'm unsure about. Say, for
instance, I have three objects in the same MTS transaction, all with a transaction
mode of Required. The first object successfully performs a SQL update, then
calls a method on the second object which also successfully performs a SQL
update then calls a method on a third object which is supposed to perform
yet another SQL update. If the third object encounters an error and I call
setabort, does this mean the two previous SQL updates are rolled back?


2.
What would actually happen if i didn't call SetComplete or SetAbort on an
object that's in a MTS package or COM+ application and a method on that object
encounters an error? Can this cause the entire application to hang? I've
seen cases where my COM+ application isn't working, I go into Component services
and I see one of my components "spinning", even
though there is no client accessing it. Stopping and restarting the application
always fixes this problem.


3.
In COM+, if my method has the Autocomplete attribute set to true, how does
it know when to commit or rollback transactions? Going back to the hypothetical
example in question 1, would it automatically rollback the transaction if
the third method encountered an error, and automatically commit the transaction
if no errors occurred? Right now that's what I'm assuming.


4.
Generally speaking, is it better to manually call SetAbort and
SetComplete myself, or use the AutoComplete feature? Why?


----

thanks in advance to anyone who can help me out

[Paul Klanderud]

Hi Paul,

You aren't too much of a newbie, because these are good questions you're
asking. I can answer your first 2 questions, but not the last 2, 'cause I
haven't worked with COM+ yet.

>1.
>I understand that to take advantage of MTS or COM+'s JIT (just in time)
activation
>and deactivation features, you can call the setabort and setcomplete methods
>of the objectcontext to deactivate the object. I've also read it's used
>to control transactions ... this is the part I'm unsure about. Say, for
>instance, I have three objects in the same MTS transaction, all with a transaction
>mode of Required. The first object successfully performs a SQL update, then
>calls a method on the second object which also successfully performs a SQL
>update then calls a method on a third object which is supposed to perform
>yet another SQL update. If the third object encounters an error and I call
>setabort, does this mean the two previous SQL updates are rolled back?

Yes, that's correct: everything in the transaction is rolled back. One of
the main ideas behind MTS and its "object context" is to provide an all-or-nothing
environment for possibly complex transactions (as in the example you cited).
As long as you are properly getting references to the existing ObjectContext
and calling SetComplete when your methods successfully complete -- and, conversely,
SetAbort in their error-handling routines -- MTS will take care of the entire
transaction for you.

So, for example, if you were putting an order into a database and running
separate SQL statements for the order items, the credit-card data, and the
customer address data, if you encountered a trappable error anywhere in any
of these routines and called SetAbort, all transactions would be rolled back.
(The beauty here is that the entire transaction is treated atomically without
the programmer having to mess with it.)

As an aside, one of the reasons you call SetComplete in secondary objects
is that it tells MTS that they can be released.


>2.
>What would actually happen if i didn't call SetComplete or SetAbort on an
>object that's in a MTS package or COM+ application and a method on that
object
>encounters an error? Can this cause the entire application to hang? I've
>seen cases where my COM+ application isn't working, I go into Component
services
>and I see one of my components "spinning", even
>though there is no client accessing it. Stopping and restarting the application
>always fixes this problem.

In certain cases this could cause your app to hang. If, for example, SetAbort
had been called somewhere along the line, you must raise an error back to
your root object (the base for the entire transaction) so it will know that
it should also call SetAbort before passing control back to the client. If
you fail to call SetAbort in the root object, MTS will still deactivate the
objects and roll back the transaction, but the root object can be left in
an unstable state. Subsequent attempts to invoke it may result in MTS errors
indicating "context aborted" or something along those lines.

Hope that helps
Paul K.


>3.
>In COM+, if my method has the Autocomplete attribute set to true, how does
>it know when to commit or rollback transactions? Going back to the hypothetical
>example in question 1, would it automatically rollback the transaction if
>the third method encountered an error, and automatically commit the transaction
>if no errors occurred? Right now that's what I'm assuming.
>
>
>4.
>Generally speaking, is it better to manually call SetAbort and
>SetComplete myself, or use the AutoComplete feature? Why?
>
>
>----
>
>thanks in advance to anyone who can help me out