MobileApp Monitoring - Delay in metrics

[reravindran]

Android Mobile Monitoring data (User Actions & Web requests) appear in Dynatrace console only after around 10 minutes they have actually made. What could be the reason for this delay ?

[skrystosik · Accepted solution]

Mobile app is not sending data on the same way as Web application. There data has to be sent almost immediately because you can always close browser. Mobile app has local storage, data are sent in chunks. If somebody is not clicking a lot it may be delay. Another thing is processing collected data by server to present them it takes 1-2 minutes most of the time.

You can use fiddler for monitoring mobile network traffic, check how often data are sent to dynatrace

Sebastian

[Patrick_H]

Mobile agent sends data with a max delay of 120 seconds. During correlation there is a timout of max 150 seconds to link service requests to it - so if the data has webrequests but no service to be linked is found this timeout also hits to ensure al necessary data is already available and we do not miss soemthing (if a service to be linked is found, then processing is faster). In worst case we then already are at 4.5 Minutes. After mobile correlation there still are some other analyzers processing the data. User actions will be visible in UI faster than Session related data, as the Session processing is again a different step. The generation of aggregated Metrics also happens in a scheduled way, so is not recalculated every second which also introduces some delay. Also the data needs to be stored to Cassandra/elasticsearch to be available across all cluster nodes.

Hope this sheds a bit more light on it.

[reravindran]

@Sebastian K. Is there a way to control the size or any other parameter of this "local storage" ? Our instrumented app is not yet in production and not heavily used. But we would like to get the metrics quicker than what we are receiving currently

[Patrick_H]

No, the send interval cannot be adjusted (but data is also flushed instantly when the app is backgrounded or terminated). See my Answer for a more detailed analysis of what happens.
There is the option do manually flush all finished agent measurements, but this is not recommended as it will only increase communication overhead won't really have a noticable speed impact on when you see the data on the cluster for the reasons described in my answer.