ADM monitoring configuration about snapshot
We have encounter an issue on our predictor report generation. We found that the instances in Data-Decision-ADM-PredictorBinningSnapshot are very unstable. Sometimes we see the instances and sometimes we did not see them at all which result in the empty predictor report. Would that be the reason we set the monitoring configuration to only the latest predictor data? (see picture)
I have also check the DDS that daystokeepdata is set to 180 days so the snapshot should not be cleaned out so quickly. I have tried to run the pzRunADMMonitoringTasks manueally and I get the intances again, but in the morning when it comes to the time the agent run the monitoring task the snapshot got cleaned out again.
There are some days the instances were kept which is the time point I have made some reponses to the IH table. But it only last for few days. So if the responses are not the lastest one the predicotr snapshot won't be created? What is the definition of latest here? And how the setup affect the monitoring activity task? Thanks ahead for the answering or any suggestions.
Pegasystems Inc.
NL
Hi @YuMeiL21,
What "latest" means is that historical data will not be kept. If you keep all data, then the table will contain the predictor snapshots for each time the snapshot process runs. This means it will contain also the historical data up to 180 days. With latest it means that the previous snapshots are deleted, and only the latest snapshot will remain.
This process should not be related to any responses on the IH table. When the snapshot process runs, it simply takes a snapshot of the adaptive models as they exist on the system at that time, and captures this in the database tables.
If you have models on your system, then it is not expected that the predictor binning table is empty after the snapshot process runs. If you do see empty tables, it might indicate a problem.
Have you been able to verify if the agent completes successfully, or if there are any exceptions in the logs related to it?
Best, Tom
Updated: 17 Sep 2024 8:06 EDT
Accenture GmbH
DE
Hi @TomVanDuist
Yes we always have models on our system and the running is alway success. And we found that one of the activity (pzGetAllModelDetails, step15) which is responsible for insert and deletion. Usually it should keep the new one and delete the old snapshots with the selection of "keep latest prediction data" . But sometimes the java delete even the new inserted ones. It looks like the defined delete time is too close to the instance's snapshot time. Since it is a pega adm service. We don't have chance to see how the method work or to change it. Please let us know how can we work on that.