I’m testing different configurations of a new server before migration and noticed speed decreasing for consecutive runs of the same job with the same configuration.
The job I’m running has 8.500 records and generates 17.000 pages. For the first run, after the server had started, it took ~16 minutes to complete the entire job. The same job began for the second time the moment the first run was completed and it took a few minutes longer to complete. The next run added a few more minutes and so on. After a few runs time required almost doubles.
From log files, I’ve concluded that the Merge engines are responsible for slower execution since the time required for all other steps is consistent over all runs, but every run tok Create Print Content plugin few extra minutes to complete.
The server has 32 vCPU and 72GB RAM and the same job was executed multiple times with different numbers or merge engines (8 to 24 merge engines). The Merge engine has a 2GB memory limit, but during execution, the Task manager does not report more than 1GB of used memory for each engine.
How can I prevent this from happening?