Highly inconsistent response times seen in PET for consecutive hits:

Our application is a Service REST.

Issue Description:

This issue is majorly seen in PET with high difference in service response time between first hit and next consecutive hits for the same request on the same node from POSTMAN.

This behaviour is reobserved/repeated in every 10 mins of the execution for every request. And it is seen on every inbound nodes (7 nodes).

When the first request is run, it takes more than 20 secs and for the next consecutive requests, its quick, takes less than 3 secs.

We tracked the service activity run time from step 1 till last step, and observed the time taken by the service’s main activity to be less than 2 secs for the same hit that shows 20 plus response time on POSTMAN.

This cannot be due to the node level data pages load time, FU assembly time or any of the logic present between step 1 and last step of the service activity because

• all our node level data pages refresh strategy is set for 1 hour.
• PEGA Alert logs that are captured by PEGA shows the max service total response time to be around 3 secs (when threshold is set 1 sec) when the response time on POSTMAN is shown 20 plus secs for the same interaction.
• The custom response time Info Forced logging that is captured in PEGA Rules logs also shows the response time to be around 3 secs.

Question for which answer is yet to find:

If the entire service activity itself is taking 2-3 secs, where is the rest of the time which is around 17-18 secs of the total 20 secs service response time shown on postman utilized?

Why there is a difference in response time for the same request between first and hit and next consecutive hits.

Why PEGA Alert logs are not capturing the 20 plus response time?