Can dynatrace see CPU Utilization % per CPU Core & CPU Utilization % per SQL Statement ?

[AskMe-Solutions]

Hi all,

I want to know ?

1.CPU Utilization % per CPU Core

ex. DB server have 8 core. I can see how during perdormance test each CPU core run.

2.CPU Utilization % per SQL Statement

ex. Which SQL statements cause heavy CPU usage?

Thank you

[Babar_Qayyum]

Hello @dynatrace A.

According to the below documentation link, database code-level based monitoring can see how much load is placed on your databases by individual services.

https://www.dynatrace.com/support/help/how-to-use-dynatrace/databases/#expand-1833more-on-code-level...

Regards,

Babar

[Krzysztof_Ziemi]

Hi,

This is correct - on the service level, we cen tell how may queries of what kind is the service issuing against the database and how much time did the database need to serve them. Which characterizes the load placed on the database by the service.

But we can't universally analyze the load on the database server, i.e. tell that specific queries took more database server CPU and others waited for disks, etc. As Darek said, this would be possible for some databases, but for others it's not practical or not feasible. We will have it for Oracle, we look to have it for MSSQL, perhaps. For other databases - the jury is still out.

Best regards

Kris

[dglugla · Accepted solution]

1. no - OneAgent is capable of tracking CPU usage on a host, but without splitting the metric by the CPU core

2. server side and Oracle only - DBInsights. Planned as a part of DB monitoring extensions, but there's no exact ETA

BR, Darek

[gurugogi]

@dglugla still we cant export CPU utilisation at CPU core level ?

[dglugla]

Sorry @gurugogi, but I'm not able to answer your question.

Maybe @Radu could help? I believe the question is about Oracle DB extension, right?

[TomásSeroteRoos]

@gurugogiNo, the current extension doesn't export CPU utilization per core.

Tying back to the original question, we now do report CPU time for the top SQL statements though.

[fstekelenburg]

1) Is kind of a pity. It shall have to come from native tools.
For instance, it would be helpful to detect single-threaded processes. Those that only use one core (and max that core out), while the system itself has plenty.