Some of code we have build generate sometimes OutOfMemory exceptions. I am now rewriting parts of it, to make it use less memory. Much of the problematic code has the same pattern. In simplified form it is like:

EntityCollection invoices = new EntityCollection(new InvoiceEntityFactory());

try
{
    adapter = DataAccessAdapterFactory.GetDataAdapter( true );


    invoices = GetInvoices(someDate);
            
    adapter.StartTransaction( IsolationLevel.ReadCommitted , "Trans-Invoices" );

    Method1(invoices);
    Method2(invoices);
        ...         

    adapter.Commit();
    return 0;
}
catch
{
    if(adapter != null && adapter.IsTransactionInProgress)
    {
        //draai alle database bewerkingen in de transactie terug
        adapter.Rollback();
    }
    throw;
}
finally
{
    if ( adapter != null )
        adapter.CloseConnection();
}

So here is the question: What happens when Method1 or Method2 raises an OutOfMemory exception? Will the adapter still be able to perform the rollback?

Should be OK, as the rollback is just doing ROLLBACK TRANS on the db server.

Could you elaborate a bit what eats up the memory or why you're getting outofmemory exceptions?

The main problem is that a lot of data is fetched. As an example, the function i am now working on fetches about 6800 rows and prefetches 2 additional collections (with a total of 145000 rows and 235000 rows). Furthermore, there can be more threads with similar amounts at the same time.

[Edit] By the way, does anybody know a method to find out how much memory is still available for the current process/thread?

wvnoort wrote:

The main problem is that a lot of data is fetched. As an example, the function i am now working on fetches about 6800 rows and prefetches 2 additional collections (with a total of 145000 rows and 235000 rows). Furthermore, there can be more threads with similar amounts at the same time.

That's indeed a lot.

Though a lot of memory is currently wasted on entity field info which should be shared among all instances. I hope to bring down memory usage a lot in v2.0.

In the meantime, try paging, if possible.

[Edit] By the way, does anybody know a method to find out how much memory is still available for the current process/thread?

Have you tried performance monitor? Ok, not from code, but you can examine a lot of counters related to .NET which can give you a clear picture what the memusage is of the current thread/appdomain etc.