Inconsistency between builtin:tech.generic.cpu.totalTime and builtin:service.cpu.time

[PierreLamonzie]

Hello,

trying to reconcile process group instance (pgi) cpu time and underlying services cpu time, I am facing inconsistency.

When builtin:tech.generic.cpu.totalTime is available, the figures are very different from the builtin:service.cpu.time.

And it looks like builtin:service.cpu.time is also quite different from the detailed CPU consumption we can access via the UI.

Does anybody have an idea/an advice about this topic.

(on premise install)

Thx and regards.

Pierre

[p_devulapalli]

builtin:tech.generic.cpu.totalTime and builtin:service.cpu.time are two different metrics . One is a Process total CPU time and the other is Service CPU time, the values returned would be different.

Details of both the metrics are explained in the below link, could please explain bit further on your query for better understanding ?

https://docs.dynatrace.com/docs/observe-and-explore/metrics-classic/built-in-metrics

[PierreLamonzie]

Hello,

thank you for your feedback, let me elaborate a few more.

Using the standard Data explorer:

* builtin:tech.generic.cpu.totalTime - filter on pgi with an entity selector fromRelationship.isProcess on a dedicated host, and I split by dt.entity.process_group_instance

** I am supposed to have the total cpu time by process of this host

** it gives me the list of process, with their total CPU time

** in the example I use (hard to share screenshots) the top value is "process 1" - 21 min

* builtin:service.cpu.time - filter on service with an entity selector fromRelationship.runsOnProcessGroupInstance where I reference the "process 1" which is well a pgi

* to be sure, I add and and clause and I filter also on a service via fromRelationship.runsOnHost, using the same host than builtin:tech.generic.cpu.totalTime

** I am supposed to have the service CPU time of all the services running in my "process 1" pgi

** it gives me well a list of services running on the pgi I selected in my entitySelector

** in this example the top value of the services is 1h 54 min

Then the CPU time of only 1 service of a pgi is much higher than the cpu time of the pgi which is hosting this service.

Hope it's clearer.

Thx and regards.

Pierre