cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 

This product reached the end of support date on March 31, 2021.

Clairification regarding elapsed time across agent tiers

shm
Inactive

For ruling out time sinks, we need to get some small clarification regarding the elapsed time measurement between multiple agent tiers.

When we have a purepath that gets started from say a klient machine, which crosses a middle tier and pursues to a backend tier.

how is the elapsed time retrieved? - is this done pr. agent?. as in - one of the agents could be running on a poorly synced clock/OS timing.

or do you recalculate the elapsed time after stitching the purepaths together with their respective local timings? - if this is the case, how do you sync up the time between systems?

As seen in the screenshot below - we can conclude that there is indeed ~120ms wasted time between the clients WCF call and the servers WCF receiving method. hence we might be dealing with either .net issues or network issues.

If there is discrepancies regarding the way your timing works between purepath correlation - could you please elaborate?

3 REPLIES 3

Zach_Jacobs
Participant

Hello, you might find these answers regarding Elapsed Time, useful.

https://answers.dynatrace.com/questions/152124/about-the-elapsed-time-in-purepath-tree-1.html

c_schwarzbauer
Dynatrace Champion
Dynatrace Champion

hi Stefan,

we do "synchronize" the time on the DT agent's machines with the time on the DT server machine including drift calculation and such. then we convert all absolute timestamps to server time before we correlate the PurePaths.

this means we're dealing pretty well with clocks that are already off and also clocks that are drifting away.

HTH, Christian

shm
Inactive

Hi, and thanks for further elaboration - but in this case it would seem that there is an issue with the way some .net remoting calls are measured - please see attached screenshot udklip.jpg

In this case you can see the clientside remoting call at time 1472ms - through a synchronous call to the server - the elapsed time jumps to 1868ms. Now if this is a synchronous call - the measurement of the calling method should have its responsetime equal to 1869-1472ms correct?

Do note, that the "handlereply" function is called at time 1477ms, (~5ms after the client call). But why are we seeing ~400ms server side time then? - the function call is done and the reply is handled.

This is the reason to my original question, it could be timedrift?, or theres a bug in the time measurement of these purepaths?