I've finally managed to implement basic DB persistence for the detailed performance data collected by Glassbox (i.e.Responses), as described in an earlier post.
However I face to problems with the persisted data, the first being that there is a lot of "rubbish". I'm actually only interested in performance related to the execution of end user requests but Glassbox collects much more data than that, for example for every JDBC call in the server (many of them being made by the server or Glassbox themselves). What would be the best approach to collect only statistics that are related to end user request? I could create/inject a new monitor into the entry points into our app (similarly to Struts/Spring MVC monitors) to mark what is an end-user request.
I thought about injecting my own monitors for the entry points that would create Responses with some specific "mark" (e.g. using response.set("isEndUserRequest, Boolean.TRUE)) and then modify the collector/persister to process only responses with that mark or their children. But I'm not sure that this is the best approach, i believe that there already must be a facility to distinguish end-user generated processing from other ones.
Thank you!
PS: The other problem is that there is many similar nested responses for a single request without enough data to see where they occured/how they differ. I'll create another post about that.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
The responses do include information about the application and the primary operation. JDBC calls won't correlate with any end-user operation so you could discount them that way, likewise you can just not include responses for applications you don't care about, like Glassbox itself.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I've finally managed to implement basic DB persistence for the detailed performance data collected by Glassbox (i.e.Responses), as described in an earlier post.
However I face to problems with the persisted data, the first being that there is a lot of "rubbish". I'm actually only interested in performance related to the execution of end user requests but Glassbox collects much more data than that, for example for every JDBC call in the server (many of them being made by the server or Glassbox themselves). What would be the best approach to collect only statistics that are related to end user request? I could create/inject a new monitor into the entry points into our app (similarly to Struts/Spring MVC monitors) to mark what is an end-user request.
I thought about injecting my own monitors for the entry points that would create Responses with some specific "mark" (e.g. using response.set("isEndUserRequest, Boolean.TRUE)) and then modify the collector/persister to process only responses with that mark or their children. But I'm not sure that this is the best approach, i believe that there already must be a facility to distinguish end-user generated processing from other ones.
Thank you!
PS: The other problem is that there is many similar nested responses for a single request without enough data to see where they occured/how they differ. I'll create another post about that.
The responses do include information about the application and the primary operation. JDBC calls won't correlate with any end-user operation so you could discount them that way, likewise you can just not include responses for applications you don't care about, like Glassbox itself.