As promised on the usergroup I hereby create this topic.
Apparently this topic seems endless and to make Software AG and the users aware of existing functionality I would like to recall bits of Ian Porter’s presentation.
At Nissan to monitor the Broker they were using SMF-data (and with - I believe - accounting data).
They used the RDC-exit to monitor the RPC-server.
However they - as I - still do not think this is enough.
What we (at Dansk Supermarked) really would like is a setup where we decides to follow a specific application comprising one or more services (defined by the trinity) and one or more users (defined by their userid) and from the broker - preferably in a grafical tool (i.e. the Control center).
Herein the developer can get a list of all request to and from the broker which falls into the above category. Note that it is not a dynamic view that gets updated, it is a historical view.
So whenever a developer has an error he can set the trace for this user in particular and se the messages transferred between the host and the client, having then a clear notion where to look for the error.
But if anyone else has other requirements please list them so Software AG has a chance to follow up on this.
Regards, Peter F. Frandsen.
Vicepresident of the EntireX Usergroup
I would like to mention that the monitoring is not the same as tracing.
The issue of Ian Porter was tracing and i reply here in this context.
If I remember it right then it should be possible to start a tracing for a specicic user and/or a specific service(s) or a combination of both. The output should then be a list of Broker verbs and the different Control blocks in a sequence from both client and server. I presume that this would create some output in the Job log.
Therefore I would like to add to the proposition that the start of this service should be autorized. I would not like to see that everyone can start this and then leaving it on until the broker restarts.
Thierry Casier
I agree with Peter that the important thing is that it is controlled from the broker, that you can filter on either user and/or service, and that it can be switched on and off online.
I also think Thierries point about autorization is a good.
I agree, it would be very helpful if we could turn on a “filtered” trace. In many cases it is not practical to turn on a trace in Production for even a short period of time because of the amount of output that an unfiltered trace produces. I would also like the ability to filter on ERROR-CODE or ERROR-CLASS. For example, all ERROR-CODE not = ‘00000000’ or ERROR-CLASS = ‘0003’.