Partial Instrumentation in COM+

[ChristopheS]

Hi,

I have a Windows service that is monitoring a queue from MQSeries. When a message arrives, a .NET transaction is created in order to get(consume) the message in a transactional way.

Then the service is calling a COM+ component that joins the transaction and runs the business logic on it.

When I try to instrument this (at COM+ level), even if I have many messages in the queues and many calls are done to COM+, but "performance Profiler" is only instrumenting the first call to COM+.

Do I do something wrong ?

Christophe S.

[Chris.Allen]

Unfortunately, I don't fully understand what you mean. A screen shot of:

ANTS showing the call tree +
ANTS showing the method grid results

will be very helpful.

Thanks!

[ChristopheS]

The architecture is as follows:

Client -> COM+ Package

In my scenario the client will do 1 call to COM+.

Internally inside COM+, the called component will make many calls (37) to a main business component. During my instrumentation, only 1 or 2 calls (even 0 sometimes) are captured.

Extra info:
- Type of instrumentation: Method-level timings; all methods (faster)
- My client is a Windows service
- Code is release version
- No PDB files deployed
- I have often "Interop.SEHException" logged by the profiler at the beginning of the execution (but don't know what this means)

[Chris.Allen]

Sounds like frustratingly, 'one-of-those-things'. i.e. instrumenting code can interfere with the usual operation (it can reveal otherwise-hidden race-conditions, change thread-scheduling order, change permissions and lead to accessviolations etc).

Does running in sampling mode give you more 9if less detailed) information?

You may find running the profiler *explicitly* as administrator helps. Otherwise, if you wanted we would analyse a memory dump and try to understand the part ANTS is playing in any exceptions...